Internet Draft                                        Mick Seaman
Expires May November 1997                                        3Com Corp.
draft-ietf-issll-802-00.txt
draft-ietf-issll-802-01.txt                          Andrew Smith
                                                 Extreme Networks
                                                     Eric Crawley
                                                       Bay
                                              Gigapacket Networks
                                                      November 1996
          Integrated Services over IEEE 802.1D/802.1p Networks

Status of this Memo

   This document is an Internet Draft.  Internet Drafts are working
   documents of the Internet Engineering Task Force (IETF), its Areas,
   and its Working Groups. Note that other groups may also distribute
   working documents as Internet Drafts.

   Internet Drafts are draft documents valid for a maximum of six
   months. Internet Drafts may be updated, replaced, or obsoleted by
   other documents at any time.  It is not appropriate to use Internet
   Drafts as reference material or to cite them other than as a "working
   draft" or "work in progress."

   Please check the I-D abstract listing contained in each Internet
   Draft directory to learn the current status of this or any other
   Internet Draft.

Abstract

This document describes the support of IETF Integrated Services over
LANs built from IEEE 802 network segments which are may be interconnected by
draft standard IEEE 8021.D [1] P802.1p switches.

It describes the practical capabilities and limitations of this
technology for supporting Controlled Load [8] and Guaranteed Service [9]
using the inherent capabilities of the relevant 802 technologies [5],[6]
etc. and the proposed 802.1p queuing features in switches. It IEEE P802.1p
[2] is a superset of the existing IEEE 802.1D bridging specification.
This document provides a functional model for the layer 3 to layer 2 and
user-to-network dialogue which supports admission control and defines
requirements for interoperability between switches. The special case of
such networks where the sender and receiver are located on the same
segment is also discussed.

This scheme is consistent with expands on the ISSLL over 802 LANs framework discussed
at described in

[7]. It makes reference to an admission control signaling protocol
developed by the October 1996 ISSLL interim meeting WG which is known as the "Subnet Bandwidth
Manager". This is an extension  to the IETF's RSVP protocol [4] and is
described in [7]. a separate document [10].

1. Introduction

The IEEE 802.1 Interworking Task Group is currently enhancing the basic
MAC Service provided in Bridged Local Area Networks (aka "switched
LANs"). As a supplement to the original IEEE MAC Bridges standard [1] , [1],
the update P802.1p
[2], [2] proposes differential traffic class queuing ("priorities") and
access to media on the basis of a "user_priority" signaled in frames.

In this document we
* review the meaning and use of user_priority in LANs and the frame
forwarding capabilities of a standard LAN switch.
* examine alternatives for identifying layer 2 traffic flows for
admission control.
* review the options available for policing traffic flows.
* derive requirements for consistent priority traffic class handling in a network
of switches and use these requirements to discuss priority queue handling
alternatives for 802.1p and the way in which these meet administrative
and interoperability goals.
* consider the benefits and limitations of this switched-based approach,
contrasting it with full router based RSVP implementation in terms of
complexity, utilisation of transmission resources and administrative
controls.

We then describe a

The model which: used is outlined in the "framework document" [7] which in
summary:
* partitions the admission control process into two separable
operations:
* an interaction between the user of the integrated service and the
local network elements ("provision of the service" in the terms of
802.1D) to confirm the availability of transmission resources for
traffic to be introduced.
* selection of an appropriate user_priority for that traffic on the
basis of the service and service parameters to be supported.
* distinguishes between the user to network interface above and the
mechanisms used by the switches ("support of the service"). These
include communication between the switches (network to network
signaling).
* describes a simple architecture for the provision and support of these
services, broken down into components with functional and interface
descriptions:
* a single "user" component: a layer-3 to layer-2 negotiation and

translation component. component for both sending and receiving, with interfaces to
other components residing in the station.
* bridge/switch processes residing in a bridge/switch to handle admission control and
mapping requests, including proposals for actual traffic mappings to
user_priority values.
* proposes identifies a set of protocol exchange primitives based on the functions
   introduced.

This document contains much background material that is used as

justification need for the approach taken. It is anticipated that much of
this material will not form a part of the final specification. signaling protocol to carry admission control
requests between devices.

It will be noted that this document is written from the pragmatic
viewpoint that there will be a widely deployed network technology and we
are evaluating it for its ability to support some or all of the defined
IETF integrated services: this approach is intended to ensure
development of a system which can provide useful new capabilities in
existing (and soon to be deployed) network infrastructure. infrastructures.

2. Goals and Assumptions

It is assumed that the network is "switch-rich": typical subnetworks that are concerned about
quality-of-service will be"switch-rich": that is to say all most
communication between end stations using integrated services support
will pass through at least one switch. Perhaps the The mechanisms and protocols
described will be trivially extensible to communicating systems on the
same shared media, but it is important not to allow problem
generalisation to complicate the practical application that we target:
the access characteristics of Ethernet and Token-Ring LANs are forcing a
trend to switch-rich topologies together along with MAC enhancements to ensure
access predictability on half-duplex switch to switch links.

It is assumed

Note that layer-3 entities, including end-stations, we illustrate most examples in this document using RSVP as an
"upper-layer" QoS signaling protocol but there are running
the actually no real
dependencies on this protocol: RSVP could be replaced by some other
dynamic protocol in support of integrated services at that layer. No or else the requests could be made by network
management or other policy entities. In any event, no extra
modifications to this the RSVP protocol are assumed.

There may be a heterogeneous mixture of switches with different
capabilities, all compliant with IEEE 802.1p, but implementing queuing
and forwarding mechanisms in a range from simple 2-queue per port,
strict priority, up to more complex multi-queue (maybe even one per-
flow) WFQ or other algorithms.

The problem is broken down into smaller independent pieces: this may
lead to sub-optimal usage of the network resources but we contend that
such benefits are often equivalent to very small improvements in network
efficiency in a LAN environment. Therefore, it is a goal that the
switches in the network operate using a much simpler set of information
than the RSVP engine in a router. In particular, it is assumed that such

switches do not need to implement per-flow queuing and policing.

One corollary is that no per-flow policing function need take place in
the switches: it
(although they might do so).

It is a fundamental part assumption of the intserv int-serv model that flows are
isolated from each other throughout their transit across a network.
Intermediate queuing queueing nodes are expected to police the traffic to ensure
that it conforms to the pre-agreed traffic flow specification. In the
architecture proposed here for mapping to layer-2, we diverge from that
assumption in the interests of simplicity: the policing function is
assumed to be implemented in the transmit schedulers of the layer-3
devices (end stations, routers): routers). In the LAN environments envisioned, it
is reasonable to assume that end stations are "trusted" to adhere to
their agreed contracts at the inputs to the network and that we can
afford to over-allocate resources at admission -control time to
compensate for the inevitable extra jitter/bunching introduced by the
switched network itself.

These divergences have some implications on the receiver heterogeneity
that can be supported and  the statistical multiplexing gains that might
have been exploited, especially for Controlled Load flows.

3. User Priority and Frame Forwarding in IEEE 802 Networks

3.1 General IEEE 802 Service Model

User_priority is a value associated with the transmission and reception
of all frames in the IEEE 802 service model: it is supplied by a the
sender which is using the MAC service. It is provided along with the
data to a receiver using the MAC service. It may or may not be actually
carried over the network:
Token-Ring/802.5 Token- Ring/802.5 carries this value (encoded
in its FC octet), basic Ethernet/802.3 does not. 802.1p defines a way to
carry this value over the network in a similar consistent way on Ethernet, Token
Ring, FDDI or other MACs MAC-layer media using an extended frame format. The "user_priority" or "traffic class" (the latter term
usage of user_priority is to be
preferred and it summarised below but is the title more fully described
in section 2.5 of the 802.1D [1] and 802.1p document) field [2] "Support of the Internal
Layer Service by Specific MAC Procedures" and readers are referred to
these documents for further information.

If the "user_priority" is carried explicitly in packets packets, its utility is
as a simple label in the data stream enabling packets in different
classes to be discriminated easily by downstream nodes. nodes without their
having to parse the packet in more detail.

Apart from making the job of desktop or wiring-closet switches easier, it
an explicit field means they do not have to change (hardware hardware or software) software
as the rules for classifying packets evolve (based (e.g. based on new protocols
or new policies). Layer-3
switches do More sophisticated layer-3 switches, perhaps deployed

towards the core of a network, can provide added value here by
performing the classification more accurately and, hence, utilising
network resources more
efficiently: efficiently or providing better protection of
flows from one another: this appears to be a good economic choice since
there are likely to be very many more desktop/wiring closet switches in
a network than switches requiring layer 3 layer-3 functionality.

The IEEE 802 specifications make no assumptions about how user_priority
is to be used by end stations or by the network, network. In particular it can
only be considered a "priority" in a loose sense: although the current
802.1p draft defines static priority queuing as the default mode of
operation of all switches that implement multiple queues (user_priority is
defined as a 3-bit quantity
with so strict priority queueing would give value
7 = high priority, 0 = low priority). The general switch algorithm in
this case is as
follows: packets are placed onto a particular queue based on the
received user_priority (from the packet if a 802.1p header or 802.5
network was used, invented according to some local policy if not). The
selection of queue is based on a mapping from user_priority
[0,1,2,3,4,5,6 or 7] onto the number of available queues - queues.Note that
switches may implement any number of queues from 1 upwards. On transmit, any/all
frames from a higher priority queue are sent first before transmitting upwards and it may
not be visible externally, except through any from advertised switch
parameters and the its admission control behaviour, which user_priority
values get mapped to the same vs. Different queues internally.Other
algorithms that a lower priority queue. switch might implement might include e.g. weighted
fair queueuing, round robin.

In particular, IEEE makes no recommendations about how a sender should
select the value for user_priority: one of the main purposes of this
draft
current document is to propose such usage rules.

Additionally, there are no IEEE 802-defined rules for switches to agree
on and how to treat frames with different user_priority values: later on in communicate
the semantics of the values between switches, end- stations and routers.
For the remainder of this draft document we make some recommendations as to what information needs to
be shared amongst switches.

4. Mapping use the term "traffic class" when
discussing the treatment of integrated services to layer-2 in layer-3 devices

The end-station or router itself is responsible for local admission
control and scheduling packets onto its link in accordance with one of the
service agreed. Just as in the intserv model, this involves per- flow
schedulers somewhere in every such data source: it user_priority
values.

3.2 Ethernet/802.3

There is an implementation
issue whether there are separate schedulers for layer-3 and layer-2 no explicit traffic class or
whether these are combined.

5. Mapping of integrated services through layer-2 switches

5.1 Queuing

Connectionless packet-based networks in general and LAN switched
networks in particular, work today because of scaling choices user_priority field carried in network
provisioning. Consciously
Ethernet packets. This means that user_priority must be regenerated at a
downstream receiver or (more usually) unconsciously, enough excess
bandwidth and buffering is provisioned switch according to some defaults or by parsing
further into higher-layer protocol fields in the network to absorb packet. Alternatively,
the IEEE 802.1Q encapsulation [11] may be used which provides an
explicit traffic sourced by higher-layer protocols or cause their transmission
windows to run out, class field on top of an basic MAC format.

For the different IP packet encapsulations used over Ethernet/802.3, it
will be necessary to adjust any admission- control calculations
according to the framing and to the padding requirements:

Encapsulation                          Framing Overhead  IP MTU
                                          bytes/pkt       bytes

IP EtherType (ip_len<=46 bytes)             64-ip_len    1500
             (1500>=ip_len>=46 bytes)         18         1500

IP EtherType over 802.1p/Q (ip_len<=42)     64-ip_len    1500*
             (1500>=ip_len>=42 bytes)         22         1500*

IP EtherType over LLC/SNAP (ip_len<=40)     64-ip_len    1492
             (1500>=ip_len>=40 bytes)         24         1492

* note that the draft IEEE 802.1Q specification exceeds the IEEE 802.3
maximum packet length values by 4 bytes.

3.3 Token-Ring/802.5

The token ring standard [6] provides a statistical basis, so priority mechanism that can be
used to control both the network is only
overloaded queuing of packets for a short duration transmission and the average expected loading is less
than 60% (usually much less).

With the advent
access of time-critical traffic such overprovisioning has
become far less easy packets to achieve. Time critical frames may find
themselves queued for annoyingly long periods the shared media. The priority mechanisms are
implemented using bits within the Access Control (AC) and the Frame
Control (FC) fields of time behind temporary
bursts a LLC frame. The first three bits of file transfer traffic, particularly at network bottleneck
points, e.g. at the 100 Mb/s to 10 Mb/s transition that might occur
between AC
field, the riser to Token Priority bits, together with the wiring closet and last three bits of the final link
AC field, the Reservation bits, regulate which stations get access to
the user ring. The last three bits of the FC field of an LLC frame, the User
Priority bits, are obtained from a desktop switch. In this case, however, if the higher layer in the user_priority
parameter when it is known (guaranteed requests transmission of a packet. This parameter also
establishes the Access Priority used by application design, merely expected on the basis of statistics, or
just that this MAC. The user_priority value
is all that conveyed end-to-end by the network guarantees to support) that User Priority bits in the
time critical traffic FC field and is a small fraction
typically preserved through Token-Ring bridges of all types. In all
cases, 0 is the total bandwidth, it
suffices lowest priority.

Token-Ring also uses a concept of Reserved Priority: this relates to give it strict priority over the "normal" traffic. The worst
case delay experienced by
value of priority which a station uses to reserve the time critical traffic is roughly token for the
maximum next
transmission time of on the ring.  When a maximum length non-time-critical frame -
less than free token is circulating, only a millisecond for 10 Mb/s Ethernet, and well below
station having an end to
end budget based on human perception times.

When more Access Priority greater than one "priority" service is or equal to be offered by a network
element e.g. it supports controlled-load as well as Guaranteed Service,
the queuing discipline becomes more complex. In order to provide the

required isolation between Reserved
Priority in the service classes, it token will probably be
necessary allowed to queue them separately. There seize the token for
transmission. Readers are referred to [14] for further discussion of
this topic.

A token ring station is theoretically capable of separately queuing each
of the eight levels of requested user priority and then an issue transmitting
frames in order of how priority.  A station sets Reservation bits according
to
service the queues - a combination user priority of admission control and maybe
weighted fair queuing may be required frames that are queued for transmission in such cases. As with the service
specifications themselves, it is not
highest priority queue.  This allows the place for this document to
specify queuing algorithms, merely access mechanism to observe ensure that
the external
behaviour meet the services' requirements.

5.2 Multicast Heterogeneity

IEEE 802.1D and 802.1p use a model for multicast whereby a switch
performs multicast routing decisions based on frame with the destination address:
this would produce a list of output ports to which highest priority throughout the packet should entire ring will be
forwarded. In its default mode, such a switch would use
transmitted before any
user_priority value in received packets lower priority frame.  Annex I to enqueue the packets at each
output port.

At layer-3, the intserv model allows heterogeneous multicast flows where
different branches of a tree can have different types of reservations
for a given multicast destination, or even supports IEEE 802.5
token ring standard recommends that stations send/relay frames as

follows:

            Application             user_priority
            non-time-critical data      0
                  -                     1
                  -                     2
                  -                     3
            LAN management              4
            time-sensitive data         5
            real-time-critical data     6
            MAC frames                  7

To reduce frame jitter associated with high-priority traffic, the notion annex
also recommends that some
trees will have some branches with reserved flows only one frame be transmitted per token and some using best
effort (default) service.

If a switch that
the maximum information field size be 4399 octets whenever delay-
sensitive traffic is selecting per-port output queues based only on traversing the
incoming user_priority, it will have to treat all branches ring.  Most existing implementations
of token ring bridges forward all
multicast sessions within that user_priority class LLC frames with the same queuing
mechanism: no heterogeneity is then possible (if it were to implement a
separate mapping at each output port then some limited form default access
priority of
heterogeneity could be supported). It is proposed 4.  Annex I recommends that per-
user_priority queuing support is adequate as minimum standard
functionality for systems *in bridges forward LLC frames that
have a LAN environment*. Layer-3 switches
(a.k.a. routers) can be used if more flexible forms of heterogeneity are
considered necessary: their behaviour is well standardised.

6. Selecting User Priority classes

One fundamental question is "who gets user priorities greater that 4 with a reservation equal to decide what the classes mean
user priority (although the draft IEEE P802.1p [2] permits network
management override this behaviour). The capabilities provided by token
ring's user and who gets access to them?" One approach would be reservation priorities and by IEEE 802.1p can provide
effective support for Integrated Services flows that request QoS using
RSVP. These mechanisms can provide, with few or no additions to the meanings of
token ring architecture, bandwidth guarantees with the classes network flow
control necessary to support such guarantees.

For the different IP packet encapsulations used over Token Ring/802.5,
it will be "well-known": we would then need necessary to standardise a set
of classes e.g. 1 = best effort, 2 = controlled- load, 3 = guaranteed
(loose delay bound, high bandwidth), 4 = guaranteed (slightly tighter
delay) etc. The values adjust any admission-control calculations
according to encode in such a table in end stations, in
isolation from the network to which they are connected, framing requirements:

Encapsulation                          Framing Overhead  IP MTU
                                          bytes/pkt       bytes

IP EtherType over 802.1p/Q                    29          4370*
IP EtherType over LLC/SNAP                    25          4370*

*the suggested MTU from RFC 1042 [13] is
problematical: the best we could probably do would be 4464 bytes but there are issues
related to define on
user_priority value per intserv service type discovering what the maximum supported MTU between any two
points both within and leave it at that
(reserving between Token Ring subnets. We recommend here an
MTU consistent with the rest 802.5 Annex I recommendation.

4. Integrated services through layer-2 switches

4.1 Summary of switch characteristics

For the combinations for future traffic classes -
there sake of illustration, we divide layer-2 bridges/switches into
several categories, based on the level of sophistication of their QoS

and software protocol capabilities: these categories are sure not intended to
represent all possible implementation choices but, instead, to aid
discussion of what QoS capabilities can be plenty!).

We propose expected from a more flexible mapping: clients ask "the network" which
user_priority network made
of these devices.

Class I  - 802.1p priority queueuing between traffic class classes.
         - No multicast heterogeneity.
         - 802.1p GARP/GMRP pruning of individual multicast addresses.

Class II As (I) plus:
         - can map received user_priority on a per-input-port basis to use for
some internal set of canonical values.
         - can map internal canonical values onto transmitted
user_priority on a given traffic flow, as
categorised by its flow-spec per-output-port basis giving some limited form of
multicast heterogeneity.
         - maybe implements IGMP snooping for pruning.

Class III As (II) plus:
         - per-flow classification
         - maybe per-flow policing and/or reshaping
         - WFQ or other transmit scheduling (probably not per-flow) 4.2
Queueing

Connectionless packet-based networks in general, and layer-2 endpoints. The LAN-switched
networks in particular, work today because of scaling choices in network provides
a value back to the requester which
provisioning. Consciously or (more usually) unconsciously, enough excess
bandwidth and buffering is appropriate to provisioned in the current network topology, load conditions, other admitted flows etc. The task of
configuring switches with this mapping (e.g. through network management to absorb the
traffic sourced by higher-layer protocols or some other switch-switch protocol) cause their transmission
windows to run out, on a statistical basis, so that the network is only
overloaded for a short duration and the average expected loading is an order of magnitude less
complex
than performing 60% (usually much less).

With the same function in end stations. Also, when
new services (or other network reconfigurations) are added to advent of time-critical traffic such a
network, the overprovisioning has
become far less easy to achieve. Time critical frames may find
themselves queued for annoyingly long periods of time behind temporary
bursts of file transfer traffic, particularly at network elements will typically be bottleneck
points, e.g. at the ones 100 Mb/s to be upgraded
with new queuing algorithms etc. 10 Mb/s transition that might occur
between the riser to the wiring closet and can be provided with new mappings
at this time.

Given the need for a new session or "flow" requiring some QoS support, a
client then needs answers final link to the following questions:

1. which traffic class do I add user
from a desktop switch. In this flow to?
    The client needs to know how to label case, however, if it is known (guaranteed
by application design, merely expected on the packets basis of statistics, or
just that this is all that the flow as it
    places them into network guarantees to support) that the network.

2. who do I ask/tell?
    The proposed model
time critical traffic is that a client ask "the network" which
    user_priority traffic class to use for a given traffic flow. This has
    several benefits as compared to a model which allows clients to select
    a class for themselves.

3. how do I ask/tell them?
    A request/response protocol is needed between client and network: in
    fact, small fraction of the request can be piggy-backed onto an admission control request
    and total bandwidth, it
suffices to give it strict priority over the response can be piggy-backed onto an admission control
    acknowledgment. "normal" traffic. The network (i.e. the first network element encountered downstream from
the client) must then answer worst
case delay experienced by the following questions:

1. which time critical traffic class do I add this flow to?
    This is roughly the
maximum transmission time of a packing problem, difficult maximum length non-time-critical frame -
less than a millisecond for 10 Mb/s Ethernet, and well below an end to
end budget based on human perception times.

When more than one "priority" service is to solve in general, but many
    simplifying assumptions can be made: presumably some simple form of
    allocation can be done without offered by a network
element e.g. it supports Controlled-Load as well as Guaranteed Service,
the queuing discipline becomes more complex scheme able complex. In order to
    dynamically shift flows around provide the
required isolation between classes.

2. which traffic class has worst-case parameters which meet the needs of
        this flow?
    This might service classes, it will probably be
necessary to queue them separately. There is then an ordering/comparison problem: which issue of two how to
service
    classes is "better" than another? Again, we can make this tractable by
    observing that all of the current intserv classes can be ranked (best
    effort <= Controlled Load <= Guaranteed Service) in queues - a simple manner. If
    any classes are implemented in the future that cannot combination of admission control and maybe
weighted fair queuing may be simply ranked
    then required in such cases. As with the issue can be finessed by either a priori knowledge about what
    classes are supported or by configuration.

and return service
specifications themselves, it is not the chosen user_priority value place for this document to the client.

Note
specify queuing algorithms, merely to observe that the client may be either an end station, router or external
behaviour meet the services' requirements.

4.3 Multicast Heterogeneity

IEEE 802.1D and 802.1p specify a basic model for multicast whereby a first
switch performs multicast routing decisions based on the destination
address: this would produce a list of output ports to which may the packet
should be acting as a proxy for forwarded. In its default mode, such a client which does not
participate switch would use any
user_priority value in these protocols for whatever reason. Note also that a
device e.g. a server or router, may choose received packets to implement both enqueue the
"client" as well as packets at each
output port. All of the "network" portion classes of switch identified above can support
this operation.

At layer-3, the int-serv model so that it allows heterogeneous multicast flows
where different branches of a tree can
select its own user_priority values: such an implementation is, however,
discouraged unless the device really does have different types of
reservations for a close tie-in with given multicast destination, or even supports the
network topology
notion that some trees will have some branches with reserved flows and resource allocation policies.

7. Flow Identification

Several previous proposals for intserv over lower-layers have treated
switches very much as
some using best effort (default) service.

If a special case of routers: in particular, that
switches along the data path will make packet handling decisions switch is selecting per-port output queues based only on the RSVP flow and filter specifications and use them to classify the
corresponding data packets. However, filtering to the per-flow level
becomes cost-prohibitive with increasing switch speed: devices with such
filtering capabilities are unlikely to have a very different
implementation cost to IP routers, in which case we
incoming user_priority, as described by 802.1p, it must question
whether a specification oriented toward switched networks is treat all
branches of any
benefit at all.

This document proposes all multicast sessions within that "flow" identification based in user_priority
be class with
the minimum required of switches.

8. Reserving Network Resources - Admission Control

So far we have not discussed admission control. In fact, without
admission control it same queuing mechanism: no heterogeneity is possible then possible.I If a
switch were to scratchbuild implement a LAN network of separate user_priority mapping at each output
port, as described under "Class II switch" above, then some
size capable limited form
of supporting real-time services, providing that the
traffic fits within certain scaling constraints (relative link speeds,
numbers receiver heterogeneity can be supported e.g. forwarding of ports etc. - see below). This is not surprising since it is
possible to run a fair approximation to real time services traffic as
user_priority 4 on small LANs
today with no one branch where receivers have performed admission
control or help from encoded priority bits.

Imagine a campus network providing dedicated 10 Mbps connections to each
user. Each floor reservations and as user_priority 0 on one where they have not.
We assume that per-user_priority queuing without taking account of each building supports up to 96 users, organized
into groups input
or output ports is the minimum standard functionality for systems in a
LAN environment (Class I switch, as defined above). More functional
layer-2 switches or even layer-3 switches (a.k.a. routers) can be used
if even more flexible forms of 24, with each group being supported by heterogeneity are considered necessary:
their behaviour is well standardised.

4.4 Override of incoming user_priority

In some cases, a 100 Mbps
downlink to network administrator may not trust the user_priority
values contained in packets from a basement switch which concentrates 5 floors (20 x 100
Mbps) source and a data center (4 x 100 Mbps) may which to a 1 Gbps link map these
into some more suitable set of values. Alternatively, due perhaps to an 8 Gbps
central campus switch, which in turn hooks 6 buildings together (with 2

x 1 Gbps full-duplex links

equipment limitations or transition periods, values may need to support be
mapped to/from different regions of a corporate server farm). Such network.

Some switches may implement such a
network could support 1.5 Mb/s function on input that maps received
user_priority into some internal set of voice/video from every user to any
other user or (for half values (this table is known in
802.1p as the population) "user_priority regeneration table"). These values can then
be mapped using the server farm, provided output table described above onto outgoing
user_priority values: these same mappings must also be used when
applying admission control to requests that use the
video ran high priority: this gives 3000 users, all with desktop video
conferencing running along with file transfer/email etc. In such a
network RSVP's role would user_priority values
(see e.g. [10]).  More sophisticated approaches may also be limited to ensuring resource availability
at the communicating end stations envisioned
where a device polices traffic flows and for connection adjusts their onward
user_priority based on their conformance to the wide area.

In such a network, a admitted traffic flow
specifications.

4.5 Remapping of non-conformant aggregated flows

One other topic under discussion as to in the best service policy to apply
to high and low priority queues may prove academic: while it int-serv context is true
that "normal" how to
handle the traffic may be delayed by bunches of high priority frames,
queuing theory tells us for data flows from sources that are exceeding their
currently agreed traffic contract with the average queue occupancy in the high
priority queue at any switch port will be somewhat network. An approach that
shows much promise is to treat such traffic with "somewhat less than 1 (with
real user behaviour, i.e. not all watching video conferences all the
time) it should be far less. A cheaper alternative
best effort" service in order to buying equipment
with a fancy queue protect traffic that is normally given
"best effort" service policy may from having to back off (such traffic is often
"adaptive" using TCP or other congestion control algorithms and it would
be unfair to buy equipment with more
bandwidth penalise it due to lower the average link utilisation badly behaved traffic from reserved
flows which are usually set up by a few per cent.

In practice a number of objections can non-adaptive applications).

A solution here might be made to such assign normal best effort traffic to one
user_priority and to label excess non-conformant traffic as a simple
solution. There may be long established expensive equipment in the
network which does not provide all the bandwidth required. There will be
considerable concern over who "lower"
user_priority. This topic is allowed to say what further discussed below.

5. Selecting traffic classes

One fundamental question is high
priority. There may be a wish "who gets to give some form decide what the classes mean
and who gets access to them?" One approach would be for the meanings of "prioritised" service
the classes to crucial business applications, above that given be "well-known": we would then need to experimental
video-conferencing. standardise a set
of classes e.g. 1 = best effort, 2 = controlled- load, 3 = guaranteed
(loose delay bound, high bandwidth), 4 = guaranteed (slightly tighter
delay) etc. The task that faces us is values to provide encode in such a degree of
control without making that control so elaborate table in end stations, in
isolation from the network to implement which they are connected, is
problematical: one approach could be to define one user_priority value
per int-serv service and leave it at that (reserving the
control oriented solution is not simply rejected in favor rest of providing
yet the
combinations for future traffic classes - there are sure to be plenty!).

We propose here a more bandwidth, at flexible mapping: clients ask "the network" which
user_priority traffic class to use for a lower cost. given traffic flow, as
categorised by its flow-spec and layer-2 endpoints. The proposed admission control mechanism requires a query-response
interaction with the network returning provides
a "YES/NO" answer and, if
successful, the user_priority value with back to the requester which is appropriate to tag the data frames current

network topology, load conditions, other admitted flows etc. The task of
configuring switches with this flow.

9. Client mapping to layer 2

We assume (e.g. through network management,
a switch-switch protocol or via some network-wide QoS-mapping directory
service) is an order of magnitude less complex than performing the same host model as intserv
function in end stations. Also, when new services (or other network
reconfigurations) are added to such a network, the network elements will
typically be the ones to be upgraded with new queuing algorithms etc.
and RSVP: can be provided with new mappings at this time.

Given the client is running
an RSVP process which presents need for a new session establishment interface or "flow" requiring some QoS support, a
client then needs answers to
applications, signals RSVP over the network, programs scheduler and
classifiers in following questions:

1. which traffic class do I add this flow to?
 The client needs to know how to label the driver and interfaces packets of the flow as it
places them into the network.

2. who do I ask/tell?
 The proposed model is that a client ask "the network" which
user_priority traffic class to use for a policy control module. In
particular, RSVP also interfaces given traffic flow. This has
several benefits as compared to a local model which allows clients to select a
class for themselves.

3. how do I ask/tell them?
 A request/response protocol is needed between client and network: in
fact, the request can be piggy-backed onto an admission control module: it
is request
and the response can be piggy-backed onto an admission control
acknowledgment: this entity that we focus on here.

The following diagram is taken from "one pass" assignment has the RSVP spec:
                      _____________________________
                     |  _______                    |
                     | |       |   _______         |
                     | |Appli- |  |       |        |   RSVP
                     | | cation|  | RSVP <--------------------
      | |       <--       |        |
                     | |       |  |process|  _____ |
                     | |_._____|  |       --Polcy||
                     |   |        |__.__._| |Cntrl||
                     |   |data       |  |   |_____||
                     |===|===========|==|==========|
                     |   |   --------|  |    _____ |
                     |   |  |        |  ----Admis||
                     |  _V__V_    ___V____  |Cntrl||
                     | |      |  |        | |_____||
                     | |Class-|  | Packet |        |
                     | | ifier|==Schedulr|====================
                     | |______|  |________|        |    data
                     |                             |
                     |_____________________________|

                      Figure 1 - RSVP in Hosts

The local admission control entity (known as "TUTU") within a client is
responsible for mapping these layer-3 requests in TO layer TwO language.

The upper-layer entity requests from TUTU:

    "May I reserve for traffic with <traffic characteristic with
      <performance requirements from <here to <there and how
      should I label it?"

where
  <traffic characteristic = Flow Spec, Tspec, Rspec (e.g.
              bandwidth, burstiness, MTU etc.)
  <performance requirements = latency, jitter bounds etc.
  <here = IP address(es)
  <there = IP address(es) - may be multicast

The TUTU entity:
* maps the endpoints benefit of completing
the conversation to layer-2 addresses in the
   LAN, so it can figure out what traffic is really going where.
* applies local admission control on outgoing link and driver (may have
   some interaction with classifier in a timely way and scheduler here e.g. reducing the exposure to give
   classifier information about
changing conditions which user_priority values to expect)
* formats a request to could occur if clients cached the knowledge
for extensive periods.

The network with the mapped addresses and flow
   specs
* receives response from (i.e. the first network and reports element encountered downstream from
the YES/NO admission
   control client) must then answer and, for successful requests, the resulting
   user_priority back following questions:

1. which traffic class do I add this flow to?
 This is a packing problem, difficult to solve in general, but many
simplifying assumptions can be made: presumably some simple form of
allocation can be done without a more complex scheme able to dynamically
shift flows around between classes.

2. which traffic class has worst-case parameters which meet the needs of
this flow?
 This might be an ordering/comparison problem: which of two service
classes is "better" than another? Again, we can make this tractable by
observing that all of the current int-serv classes can be ranked (best
effort <= Controlled Load <= Guaranteed Service) in a simple manner. If
any classes are implemented in the future that cannot be simply ranked
then the issue can be finessed by either a priori knowledge about what

classes are supported or by configuration.

and return the chosen user_priority value to the client.

Note that the client may be either an end station, router or a first
switch which may be acting as a proxy for a client which does not
participate in these protocols for whatever reason. Note also that a
device e.g. a server or router, may choose to implement both the
"client" as well as the "network" portion of this model so that it can
select its own user_priority values: such an implementation would,
however, be discouraged unless the device really does have a close tie-
in with the network topology and resource allocation policies but would
work in some cases where there is known over- provisioning of resources.

6. Flow Identification

Several previous proposals for int-serv over lower-layers have treated
switches very much as a special case of routers: in particular, that
switches along the data path will make packet handling decisions based
on the RSVP flow and filter specifications and use them to classify the
corresponding data packets. However, filtering to the per-flow level
becomes cost-prohibitive with increasing switch speed: devices with such
filtering capabilities are unlikely to have a very different
implementation cost to IP routers, in which case we must question
whether a specification oriented toward switched networks is of any
benefit at all.

This document proposes that "aggregated flow" identification based on
user_priority be the minimum required of switches.

7. Reserving Network Resources - Admission Control

So far we have not discussed admission control. In fact, without
admission control it is possible to scratchbuild a LAN network of some
size capable of supporting real-time services, providing that the
traffic fits within certain scaling constraints (relative link speeds,
numbers of ports etc. - see below). This is not surprising since it is
possible to run a fair approximation to real time services on small LANs
today with no admission control or help from encoded priority bits.

Imagine a campus network providing dedicated 10 Mbps connections to each
user. Each floor of each building supports up to 96 users, organized
into groups of 24, with each group being supported by a 100 Mbps
downlink to a basement switch which concentrates 5 floors (20 x 100
Mbps) and a data center (4 x 100 Mbps) to a 1 Gbps link to an 8 Gbps
central campus switch, which in turn hooks 6 buildings together (with 2

x 1 Gbps full duplex links to support a corporate server farm). Such a
network could support 1.5 Mb/s of voice/video from every user to any
other user or (for half the population) the server farm, provided the
video ran high priority: this gives 3000 users, all with desktop video
conferencing running along with file transfer/email etc. In such a
network RSVP's role would be limited to ensuring resource availability
at the communicating end stations and for connection to the wide area.

In such a network, a discussion as to the best service policy to apply
to high and low priority queues may prove academic: while it is true
that "normal" traffic may be delayed by bunches of high priority frames,
queuing theory tells us that the average queue occupancy in the high
priority queue at any switch port will be somewhat less than 1 (with
real user behaviour, i.e. not all watching video conferences all the
time) it should be far less. A cheaper alternative to buying equipment
with a fancy queue service policy may be to buy equipment with more
bandwidth to lower the average link utilisation by a few per cent.

In practice a number of objections can be made to such a simple
solution. There may be long established expensive equipment in the
network which does not provide all the bandwidth required. There will be
considerable concern over who is allowed to say what traffic is high
priority. There may be a wish to give some form of "prioritised" service
to crucial business applications, above that given to experimental
video-conferencing. The task that faces us is to provide a degree of
control without making that control so elaborate to implement that the
control-oriented solution is not simply rejected in favor of providing
yet more bandwidth, at a lower cost.

The proposed admission control mechanism requires a query-response
interaction with the network returning a "YES/NO" answer and, if
successful, a user_priority value with which to tag the data frames of
this flow.

The relevant int-serv specifications describe the parameters which need
to be considered when making an admission control decision at each node
in the network path between sender and receiver. We discuss how to
calculate these parameters for different network technologies below but
we do not specify admission control algorithms or mechanisms as to how
to progress the admission control process across the network. One such
mechanism is described as SBM in [10].

Where there are multiple mechanisms in use for allocating resources e.g.
some combination of SBM and network management, it will be necessary to
ensure that network resources are partitioned amongst the different
mechanisms in some way: this could be by configuration or maybe by
having the mechanisms allocate from a common resource pool within any
device.

8. Mapping of integrated services to layer-2 in layer-3 devices

8.1 Layer-3 client

We assume the same client model as int-serv and RSVP where we use the
term "client" to mean the entity handling QoS in the layer-3 device at
each end of a layer-2 hop (e.g. end-station, router). The sending client
itself is responsible for local admission control and scheduling packets
onto its link in accordance with the service agreed. Just as in the
int-serv model, this involves per-flow schedulers (a.k.a. shapers) in
every such data source.

The client is running an RSVP process which presents a session
establishment interface to applications, signals RSVP over the network,
programs a scheduler and classifier in the driver and interfaces to a
policy control module. In particular, RSVP also interfaces to a local
admission control module: it is this entity that we focus on here.

The following diagram is taken from the RSVP specification [4]:
                      _____________________________
                     |  _______                    |
                     | |       |   _______         |
                     | |Appli- |  |       |        |   RSVP
                     | | cation|  | RSVP <-------------------->
                     | |       <-->       |        |
                     | |       |  |process|  _____ |
                     | |_._____|  |       -->Polcy||
                     |   |        |__.__._| |Cntrl||
                     |   |data       |  |   |_____||
                     |===|===========|==|==========|
                     |   |   --------|  |    _____ |
                     |   |  |        |  ---->Admis||
                     |  _V__V_    ___V____  |Cntrl||
                     | |      |  |        | |_____||
                     | |Class-|  | Packet |        |
                     | | ifier|==>Schedulr|====================>
                     | |______|  |________|        |    data
                     |                             |
                     |_____________________________|

                    Figure 1 - RSVP in Sending Hosts

Note that we illustrate examples in this document using RSVP as the
"upper-layer" signaling protocol but there are no actual dependencies on
this protocol: RSVP could be replaced by some other dynamic protocol or
else the requests could be made by network management or other policy
entities.

8.2 Requests to layer-2

The local admission control entity within a client is responsible for
mapping these layer-3 requests into layer-2 language.

The upper-layer entity requests from ISSLL:

"May I reserve for traffic with <traffic characteristic> with
<performance requirements> from <here> to <there> and how
should I label it?"

where
  <traffic characteristic> = Flow Spec, Tspec, Rspec (e.g.
              bandwidth, burstiness, MTU etc.)
  <performance requirements> = latency, jitter bounds etc.
  <here> = IP address(es)
  <there> = IP address(es) - may be multicast

8.3 Sender

The ISSLL  functionality in the sender is illustrated below and may be
summarised as:
* maps the endpoints of the conversation to layer-2 addresses in the
LAN, so it can figure out what traffic is really going where (probably
makes reference to the ARP protocol cache for unicast or an algorithmic
mapping for multicast destinations).
* applies local admission control on outgoing link and driver
* formats a SBM request to the network with the mapped addresses and
filter/flow specs
* receives response from the network and reports the YES/NO admission
control answer back to the upper layer entity, along with any negotiated
modifications to the session parameters.
* stores any resulting user_priority to be associated with this session
in a "802 header" lookup table for use when sending any future data
packets.
                    from IP     from RSVP
                   ____|____________|____________
                  |    |            |            |
                  |  __V____     ___V___         |
                  | |       |   |       |        |
                  | | Addr  |<->|       |        | SBM signaling
                  | |mapping|   | SBM   |<------------------------>
                  | |_______|   |Client |        |
                  |  ___|___    |       |        |
                  | |       |<->|       |        |
                  | |  802  |   |_______|        |
                  | | header|     / | |          |
                  | |_______|    /  | |          |
                  |    |        /   | |   _____  |
                  |    | +-----/    | +->|Local| |
                  |  __V_V_    _____V__  |Admis| |
                  | |      |  |        | |Cntrl| |
                  | |Class-|  | Packet | |_____| |
                  | | ifier|==>Schedulr|======================>
                  | |______|  |________|         |  data
                  |______________________________|

                Figure 2 - ISSLL in End-station Sender

ISSLL manageable objects in the sender:
  802 header table
  Local admission control resource status
  L2 additions to classifier/scheduler int-serv tables

8.4 Receiver

The ISSLL functionality in the receiver is a good deal simpler. It is
summarised below and is illustrated by the following picture:
* handles any received SBM protocol indications.
* applies local admission control to see if a request can be supported
with appropriate local receive resources.
* passes indications up to RSVP if OK.
* accepts confirmations from RSVP and relays them back via SBM signaling
towards the requester.
* may program a receive classifier and scheduler, if any is used, to
identify traffic classes of received packets and accord them appropriate
treatment e.g. reserve some buffers for particular traffic classes.
* programs receiver to strip any 802 header information from received
packets.

                     to RSVP       to the upper layer entity.

                    from IP
                       ^            ^
                   ____|____________|___________
                  |    |            |           |
                  |  __|____        |           |
                  | |       |       |           |
 SBM signaling    | |  SBM  |    ___|___        |
<-----------------> |Client |   | Strip |       |
                  | |_______|   |802 hdr|       |
                  |    |   \    |_______|       |
                  |  __v___ \       ^           |
                  | | Local |\      |           |
                  | | Admis | \     |           |
                  | | Cntrl |  \    |           |
                  | |_______|   \   |           |
                  |  ______     v___|____       |
                  | |Class-|   | Packet  |      |
===================>| ifier|==>|Scheduler|      |
     data         | |______|   |_________|      |
                  |_____________________________|

                Figure 3 - ISSLL in End-station Receiver

9. Layer-2 Switch Functions

9.1 Switch Model

In this model of layer-2 switch behaviour, we define the following
entities within the switch:

* Local admission control - one of these on each port accounts for the

available bandwidth on the link attached to that port. For half-duplex
links, this involves taking account of the resources allocated to both
transmit and receive flows. For full-duplex, the input port accountant's
task is trivial.

* Input SBM module: one instance on each port, performs the "network"
side of the signaling protocol for peering with clients or other
switches. Also holds knowledge of the mappings of int-serv classes to
user_priority.

* SBM propagation - relays requests that have passed admission control
at the input port to the relevant output ports' SBM modules. This will
require access to the switch's forwarding table (layer-2 "routing table"
- cf. RSVP model) and port spanning-tree states.

* Output SBM module - forwards requests to the next layer-2 or -3
network hop.

* Classifier, Queueing and Scheduler - these functions are basically as
described by the Forwarding Process of IEEE 802.1p (see section 3.7 of
[2]). The Classifier module identifies the relevant QoS information from
incoming packets and uses this, together with the normal bridge
forwarding database, to decide to which output queue of which output
port to enqueue the packet. In Class I switches, this information is the
"regenerated user_priority" parameter which has already been decoded by
the receiving MAC service and potentially re-mapped by the 802.1p
forwarding process (see description in section 3.7.3 of [2]). This does
not preclude more sophisticated classification rules which may be
applied in more complex Class III switches e.g. matching on individual
int-serv flows.

 The Queueing and Scheduler module holds the output queues for ports and
provides the algorithm for servicing the queues for transmission onto
the output link in order to provide the promised int-serv service.
Switches will implement one or more output queues per port and all will
implement at least a basic strict priority dequeueing algorithm as their
default, in accordance with 802.1p.

* Ingress traffic class mapper and policing - as described in 802.1p
section 3.7. This optional module may check on whether the data within
traffic classes are conforming to the patterns currently agreed:
switches may police this and discard or re-map packets. The default
behaviour is to pass things through unchanged.

* Egress traffic class mapper - as described in 802.1p section 3.7. This
optional module may apply re-mapping of traffic classes e.g. on a per-
output port basis. The default behaviour is to pass things through
unchanged.

These are shown by the following diagram which is a superset of the IEEE
802.1D/802.1p bridge model:

                   _______________________________
                  |  _____     ______     ______  |
 SBM signaling    | |     |   |      |   |      | | SBM signaling
<------------------>| IN  |<->| SBM  |<->| OUT  |<---------------->
                  | | SBM |   | prop.|   | SBM  | |
                  | |_____|   |______|   |______| |
                  |  / |          ^     /     |   |
    ______________| /  |          |     |     |   |_____________
   | \             / __V__        |     |   __V__             / |
   |   \      ____/ |Local|       |     |  |Local|          /   |
   |     \   /      |Admis|       |     |  |Admis|        /     |
   |       \/       |Cntrl|       |     |  |Cntrl|      /       |
   |  _____V \      |_____|       |     |  |_____|    / _____   |
   | |traff |  \               ___|__   V_______    /  |egrss|  |
   | |class |    \            |Filter| |Queue & | /    |traff|  |
   | |map & |=====|==========>|Data- |=| Packet |=|===>|class|  |
   | |police|     |           |  base| |Schedule| |    |map  |  |
   | |______|     |           |______| |________| |    |_____|  |
   |____^_________|_______________________________|______|______|
data in |                                                |data out
========+                                                +========>
                Figure 4 - ISSLL in Switches

9.2 Admission Control

On reception of an admission control request, a switch performs the
following actions:
* ingress SBM module translates any received user_priority or else
selects a layer-2 traffic class which appears compatible with the
request and whose use does not violate any administrative policies in
force. In effect, it matches up the requested service with those
available in each of the user_priority classes and chooses the "best"
one. It ensures that, if this reservation is successful, the selected
value is passed back to the client.
* ingress SBM observes the current state of allocation of resources on
the input port/link and then determines whether the new resource
allocation from the mapped traffic class would be excessive. The request
is passed to the reservation propagator if accepted so far.
* reservation propagator relays the request to the bandwidth accountants
on each of the switch's outbound links to which this reservation would
apply (implied interface to routing/forwarding database).
* egress bandwidth accountant observes the current state of allocation
of queueing resources on its outbound port and bandwidth on the link
itself and determines whether the new allocation would be excessive.
Note that this is only the local decision of this switch hop: each

further layer-2 hop through the network gets a chance to veto the
request as it passes along.
* the request, if accepted by this switch, is then passed on down the
line on each output link selected. Any user_priority described in the
forwarded request must be translated according to any egress mapping
table.

* if accepted, the switch must notify the client of the user_priority to
use for packets belonging to this flow.  Note that this is a
"provisional YES" - we assume an optimistic approach here: later
switches can still say "NO" later.
* if this switch wishes to reject the request, it can do so by notifying
the original client (by means of its layer-2 address).

10. Mappings from RSVP
                   ____|____________|____________
                  |    |      |     |            |
                  |  __V____  |  ___V___         |
                  | |       | | |       |        |
                  | |  ARP  | | |       |        | ISSLL signaling
                  | |protocl| | | TUTU  |<------------------------
                  | |       |<-|       |        |
                  | |       | | |       |        |
                  | |_______| | |       |        |
                  |    |      | |_______|        |
                  |    |data  |    |  |          |
                  |====|===========|==|==========|
                  |    |  +--------|  |   _____  |
                  |    |  |        |  +-|Local| |
                  |  __V__V_   ____V___  |Admis| |
                  | |      |  |        | |Cntrl| |
                  | |Class-|  | Packet | |_____| |
                  | | ifier|==Schedulr|======================
                  | |______|  |________|         |  data
                  |                              |
                  |______________________________|

                Figure 2 intserv service models to IEEE 802

It is assumed that admission control will be applied when deciding
whether or not to admit a new flow through a given network element and
that a device sending onto a link will be proxying the parameters and
admission control decisions on behalf of that link: this process will
require the device to be able to determine (by estimation, measurement
or calculation) several parameters. It is assumed that details of the
potential flow are provided to the device by some means (e.g. a
signaling protocol, network management). The service definition
specifications themselves provide some implementation guidance as to how
to calculate some of these quantities.

The accuracy of calculation of these parameters may not be very
critical: indeed it is an assumption of this model's being used with
relatively simple Class I switches that they merely provide values to
describe the device and admit flows conservatively.

10.1 General characterisation parameters

There are some general parameters that a device will need to use and/or
supply for all service types:
  - Ingress link
  - Egress links and their MTUs, framing overheads and minimum packet
sizes (see media-specific information presented above).
  - available path bandwidth: updated hop-by-hop by any device along the
path of the flow.
  - minimum latency

10.2 Parameters to implement Guaranteed Service

A network element must be able to determine the following parameters:

  - Constant delay bound through this device (in addition to any value
provided by "minimum latency" above) and up to the receiver at the next
network element for the packets of this flow if it were to be admitted:
this would include any access latency bound to the outgoing link as well
as propagation delay across that link.
  - Rate-proportional delay bound through this device and up to the
receiver at the next network element for the packets of this flow if it
were to be admitted.
  - Receive resources that would need to be associated with this flow
(e.g. buffering, bandwidth) if it were to be admitted and not suffer
packet loss if it kept within its supplied Tspec/Rspec.
  - ISSLL in Hosts

10. Switch Functions

10.1 Admission Control

For the sake of Transmit resources that would need to be associated with this discussion, we define flow
(e.g. buffering, bandwidth, constant- and rate-proportional delay
bounds) if it were to be admitted.

10.3 Parameters to implement Controlled Load

A network element must be able to determine the following entities within parameters
which can be extracted from [8]:

  - Receive resources that would need to be associated with this flow
(e.g. buffering) if it were to be admitted.
  - Transmit resources that would need to be associated with this flow
(e.g. buffering) if it were to be admitted.

10.4 Parameters to implement Best Effort

For a layer-2 switch:
* traffic class network element to implement best effort service there are no
explicit parameters that need to be characterised.

10.5 Mapping to IEEE 802 user_priority

There are many options available for mapping authority - aggregations of flows
described by int-serv service models (Best Effort, Controlled Load, and
Guaranteed are the services considered here) onto user_priority classes.
There currently exists very little practical experience with particular
mappings to help make a determination as to the "best" mapping.  In that
spirit, the following options are presented in order to stimulate
experimentation in this holds area. Note, this does not dictate what
mechanisms/algorithms a network element (e.g. an Ethernet switch) needs
to perform to implement these mappings: this is an implementation choice
and does not matter so long as the mapping table requirements for the particular
service model are met. Having said that, we do explore below the ability
of
   intserv classes a switch implementing strict priority queueing to user_priority.
* reservation accountants - one support some or all
of these on each port accounts for the
   available bandwidth on that link. For half-duplex links, service types under discussion: this
   involves taking account of both transmit and receive flows. For
   full-duplex the input port accountant's task is trivial.
* reservation propagators - these propagate requests that have passed
   admission control at the input port's accountant worthwhile because this
is likely to be the relevant
   output ports' accountants. This will require access most widely deployed dequeueing algorithm in simple
switches as it is the default specified in 802.1p.

In order to reduce the switch's
   forwarding administrative problems , such a mapping table (layer-2 "routing table" - cf. RSVP model) is
held by *switches* (and routers if desired) but generally not by end-
station hosts and
   spanning-tree state.

These is a read-write table. The values proposed below are shown
defaults and can be overridden by the following diagram:
                   _______________________________
                  |  _____     ______     _____   |
                  | |Span |   |filter|   |traff|  |
                  | |Tree |<-|data- |   |class|  |
                  | |Prot.|   |  base|   |map  |  |
                  | |_____|   |______|   |_____|  |
                  |              ^                |
                  |  _____     __|___     ______  |
 ISSLL signaling  | | in  |   |      |   | out  | | ISSLL signaling
<------------------|resv |<-| resv |<-| resv |<----------------
                  | |acct.|   | prop.|   | acct.| |
                  | |_____|   |______|  /|______| |
                  |    |   \           /     |    |
                  |====|====\=========|======|====|
                  |  __V__  |         |    __V__  |
                  | |Local| |         |   |Local| |
                  | |Admis| |         |   |Admis| |
                  | |Cntrl| |         |   |Cntrl| |
                  | |_____| |         |   |_____| |
                  |     ____V_      __V____       |
                  |    |Class-|    | Packet |     |
      ===============-| ifier|====Schedulr|===================
         data     |    |______|    |________|     |  data
                  |                               |
                  |_______________________________|

                    Figure 3 - ISSLL in Switches

On reception management control so long as all
switches agree to some extent (the required level of agreement requires
further analysis).

It is possible that some form of an admission control request, a switch performs the
following actions:
* ingress bandwidth accountant observes the current state network-wide lookup service could be
implemented that serviced requests from clients e.g. traffic_class =
getQoSbyName("H.323 video") and notified switches of allocation what sorts of resources on the input port/link
traffic categories they were likely to encounter and then determines whether the
   new allocation would be excessive. The request is passed how to allocate
those requests into traffic classes: such mechanisms are for further
study.

Proposal:  A Simple Scheme

     user_priority      Service
       0                "less than" Best Effort
       1                Best Effort
       2                reserved
       3                reserved
       4                Controlled Load
       5                Guaranteed Service, 100ms bound
       6                Guaranteed Service, 10ms bound
       7                reserved

In this proposal, all traffic that uses the
   reservation propagator if accepted so far.
* reservation propagator relays the request controlled load service is
mapped to the bandwidth accountants
   on each a single 802.1p user_priority whilst that for guaranteed
service is placed into one of the switch's outbound links to which this reservation
   would apply (implied interface two user_priority classes with different
delay bounds. Unreserved best effort traffic is mapped to routing/forwarding database).
* egress bandwidth accountant observes another.

The use of classes 4, 5 and 6 for Controlled Load and Guaranteed Service
is somewhat arbitrary as long as they are increasing. Any two classes
greater than Best Effort can be used as long as GS is "greater" than CL
although those proposed here have the advantage that, for transit
through 802.1p switches with only two-level strict priority queuing,
they both get "high priority" treatment (the current state 802.1p default
split is 0-3 and 4-7 for a device with 2 queues). The choice of delay
bound is also arbitrary but potentially very significant: this can lead
to a much more efficient allocation of queueing resources on its outbound port and bandwidth on the link
   itself and determines whether the new allocation would as well as greater
(though still not very good) isolation between flows.

The "less than best effort" class might be excessive.
   Note useful for devices that this is only the local decision of this switch hop: each
   further layer-2 hop through the wish
to tag packets that are exceeding a committed network gets capacity and can
be optionally discarded by a chance to veto the
   request as it passes along.
* the request, if accepted downstream device.  Note, this is not
*required* by any current int-serv models but is under study.

The advantage to this switch, approach is then passed that it puts some real delay bounds on down

the
   line on each output link selected.
* if this is Guaranteed Service without adding any additional complexity to the first switch in line,
other services.  It still ignores the amount of *bandwidth* available
for each class. This should behave reasonably well as long as all
traffic class mapping
   authority selects a layer-2 traffic class which appears compatible
   with the request for CL and whose use GS flows does not violate exceed any administrative
   policies resource capacities in force. In effect, it matches up
the requested service with
   those available device. Some isolation between very delay-critical GS and less
critical GS flows is provided but there is still an overall assumption
that flows will in each of general be well- behaved. In addition, this mapping
still leaves room for future service models.

Expanding the user_priority number of classes for CL service is not as appealing since
there is no need to map to a particular delay bound.  There may be cases
where an administrator might map CL onto more classes and chooses the
   "best" one. for particular
bandwidths or policy levels.  It ensures that, if this reservation is successful, the
   selected value is passed back may also be desirable to further
subdivide CL traffic in cases where the client.
* if accepted, the switch must notify itis frequently non-conformant
for certain applications.

11. Network Topology Scenarios

11.1 Switched networks using priority scheduling algorithms

In general, the client int-serv standards work has tried to avoid any
specification of scheduling algorithms, instead relying on implementers
to deduce appropriate algorithms from the user_priority service definitions and on
users to
   use apply measurable benchmarks to check for packets belonging conformance. However,
since one standards' body has chosen to this flow.  Note that this is specify a
   "provisional YES" - we assume an optimistic approach here: later single default
scheduling algorithm for switches can still say "NO" later.
* if this switch wishes to reject the request, [2], it can do so by notifying
   the original client (by means of its layer-2 address).

10.2 Mappings seems appropriate to IEEE 802 user_priority

There are several options available for mapping service models (Best
Effort, Controlled Load, and Guaranteed) examine
to IEEE 802.1p user_priority
classes.  The problem with making choices at some degree, how well this time is that we don't
have much experience with any particular "implementation" might actually support
some or all of the int-serv services.

If the mappings to help make of Proposal A above are applied in a
determination as switch implementing
strict priority queueing between the 8 traffic classes (7 = highest)
then the result will be that all Guaranteed Service packets will be
transmitted in preference to any other service. Controlled Load packets
will be transmitted next, with everything else waiting until both of
these queues are empty. If the admission control algorithms in use on
the switch ensure that the "best" mapping. So, sum of the following options "promised" bandwidth of all of the
GS and CL sessions are
presented to stimulate discussion in this area.  Note, this does not
dictate what mechanisms/algorithms a network element (e.g. an Ethernet
switch) needs never allowed to do implement these mappings: this is an implementation
choice and does not matter so long as the requirements for exceed the
particular service model available link
bandwidth then things are met.

In order looking good.

11.2 Full-duplex switched networks

We have up to reduce now ignored the administrative problems of maintaining such
mappings, such MAC access protocol. On a mapping table is held full-duplex
switched LAN (of either Ethernet or Token-Ring types - the MAC algorithm
is, by *switches* only (and routers
if desired) and is a read-write table. The values proposed below are
defaults and definition, unimportant) this can be overridden by management control so long as all
switches agree to some extent (the required level of agreement requires
further thought).

Option A:  The Simple Method

In this method, all traffic that uses a particular service model is
mapped factored in to a single 802.1p user_priority.  This the
characterisation parameters advertised by the device since the access
latency is fine as long as all
traffic for a given service model does not exceed any capacity well controlled (jitter = one largest packet time). Some
example characteristics (approximate):

     Type        Speed         Max Pkt   Max Access
                               Length    Latency

     Ethernet    10Mbps         1.2ms     1.2ms
                 100Mbps        120us     120us
                 1Gbps           12us      12us
     Token-Ring  4Mbps            9ms       9ms
                 16Mbps           9ms       9ms
     FDDI        100Mbps        360us     8.4ms

These delays should be also be considered in the
802 device and fine control context of delay is not needed.  Here is an example:

      Priority  Service
        0       "less than" Best Effort
        1       Best Effort
        2       reserved
        3       reserved
        4 speed- of-
light delays of e.g. ~400ns for typical 100m UTP links and ~7us for
typical 2km multimode fibre links.

Therefore we see Full-Duplex switched network topologies as offering
good QoS capabilities for both Controlled Load
        5       reserved
        6 and Guaranteed Service.

11.3 Shared-media Ethernet networks

We have not mentioned the difficulty of dealing with allocation on a
single shared CSMA/CD segment: as soon as any CSMA/CD algorithm is
introduced then the ability to provide any form of Guaranteed Service
        7       reserved

The "less than" best effort service is useful for devices that wish to
tag packets that
seriously compromised in the absence of any tight coupling between the
multiple senders on the link. There are exceeding a committed network capacity and can be
optionally discarded by number of reasons for not
offering a downstream device.  Note, better solution for this is issue.

Firstly, we do not
necessarily incorporated in any current IntServ model.

The advantage of believe this mapping is that a truly solvable problem: it leaves room for future service
models.  The choices would
seem to require a new MAC protocol. Those who are interested in solving
this problem per se should probably be following the BLAM developments
in 802.3 but we would be suspicious of priority 4 the interoperability
characteristics of a series of new software MACs running above the
traditional 802.3 MAC.

Secondly, we are not convinced that it is really an interesting problem.
While not everyone in the world is buying desktop switches today and priority 6
there will be end stations living on repeated segments for Controlled Load some time to
come, the number of switches is going up and Guaranteed Service, respectively, the number of stations on
repeated segments is somewhat arbitrary.  Any two
priorities greater than Best Effort can be used as long as Guaranteed
Service going down. This trend is "greater" than Controlled Service although those proposed
here have proceeding to the advantage that, for transit point
that we may be happy with a solution which assumes that any network
conversation requiring resource reservations will take place through 802.1p switches with
only two-level strict priority queuing, they both get "high priority"
treatment (the current 802.1p split at
least one switch (be it layer-2 or layer-3). Put another way, the
easiest QoS upgrade to a layer-2 network is 0-3 and 4-7 for 2 queues).

One disadvantage to this mapping install segment
switching: only when has been done is that it ignores worthwhile to investigate more
complex solutions involving admission control.

Thirdly, in the delay
characteristics core of the guaranteed service and groups network (as opposed to at the edges), there
does not seem to be enough economic benefit for repeated segment
solutions as opposed to switched solutions. While repeated solutions

*may* be 50% cheaper, their cost impact on the entire network is
amortised across all guaranteed
traffic, no matter what of the delay bound, into edge ports. There may be special
circumstances in the future (e.g. Gigabit buffered repeaters) but these
have differing characteristics to existing CSMA/CD repeaters anyway.

     Type        Speed         Max Pkt   Max Access
                                  Length    Latency

     Ethernet    10Mbps         1.2ms    unbounded
                 100Mbps        120us    unbounded
                 1Gbps           12us    unbounded

11.4 Half-duplex switched Ethernet networks

Many of the same priority.

Option B:  Two Classes arguments for sub-optimal support of Guaranteed Service

For
apply to half-duplex switched Ethernet as to shared media: in essence,
this method, we expand topology is a medium that *is* shared between at least two senders
contending for each packet transmission opportunity. Unless these are
tightly coupled and cooperative then there is always the number chance that the
junk traffic of priorities assigned one will interfere with the other's important traffic.
Such coupling would seem to need some form of modifications to the
Guaranteed Service:
      Priority  Service
        0       "less than" Best Effort
        1       Best Effort
        2       reserved
        3       reserved
        4 MAC
protocol (see above).

Notwithstanding this, these topologies do seem to offer the chance to
provide Controlled Load
        5       Guaranteed Service, 100ms bound
        6       Guaranteed Service, 10ms bound
        7       reserved

Again, the choices of service: with the exact priorities knowledge that there are somewhat arbitrary as
long as they only
a small limited number (e.g. two) of potential senders that are increasing.  Similarly, both
using prioritisation for their CL traffic (with admission control for
those CL flows based on the choice knowledge of delay bound is
also arbitrary but potentially very significant.  One the number of potential
senders) over best effort, the key
differences is that now there media access characteristics, whilst not
deterministic in the true mathematical sense, are somewhat predictable.
This is probably a bound on delay through the network
(and hence through each device) which may be much harder close enough approximation to implement
although it can lead CL to be useful.

     Type        Speed          Max Pkt   Max Access
                                Length    Latency

     Ethernet    10Mbps           1.2ms   unbounded
                 100Mbps          120us   unbounded
                 1Gbps             12us   unbounded

11.5 Half-duplex and shared Token Ring networks

In a much more efficient allocation of resources.

The advantage to this approach is that it puts some real delay bounds on shared Token Ring network, the Guaranteed Service without adding network access time for high
priority traffic at any additional complexity to the
other services.  It still ignores station is bounded and is given by (N+1)*THTmax,
where N is the amount number of *bandwidth* available
for each class.

Further derivations stations sending high priority traffic and
THTmax is the maximum token holding time [14]. This assumes that network
adapters have priority queues so that reservation of this option could be made by dividing the

Guaranteed Service classes into more levels token is done
for traffic with particular delay
bounds.  Expanding the number of priorities highest priority currently queued in the adapter.
It is easy to see that access times can be improved by reducing N or

THTmax.  The recommended default for Controlled Load service THTmax is not as appealing since there 10 ms [6]. N is no need to map an
integer from 2 to 256 for a particular delay
bound.  There may be shared ring and 2 for a cases where an administrator might map Controlled
Load switched half duplex
topology. A similar analysis applies for FDDI. Using default values
gives:

     Type        Speed              Max Pkt   Max Access
                                    Length    Latency

     Token-Ring  4/16Mbps shared       9ms    2570ms
                 4/16Mbps switched     9ms      30ms
     FDDI        100Mbps             360us       8ms

Given that access time is bounded, it is possible to more priorities provide an upper
bound for particular bandwidths or policy levels. It
may also be necessary to further classify Controlled Load end-to-end delays as required by Guaranteed Service assuming
that traffic in
cases and where of this class uses the Controlled Load traffic is frequently non-conformant highest priority allowable for certain applications.

10.3 Policy

A policy agent user
traffic.  The actual number of stations that send traffic mapped into
the same traffic class as GS may also be implemented by a switch. This determines, how
to interpret received user_priority values vary over time but, from packets, whether to
trust them and whether to map them to something else. an admission
control standpoint, this value is needed a priori.  The policies in
force admission
control entity must therefore use a fixed value for N, which may be configured by network management. Default the
total number of stations on the ring or some lower value if it is
desired to use what keep the offered delay guarantees smaller. If the value of N
used is
received and pass it lower than the total number of stations on unchanged.

11. Signaling protocol

It is not the intention ring, admission
control must ensure that the number of stations sending high priority
traffic never exceeds this number. This approach allows admission
control to precisely define a protocol estimate worst case access delays assuming that all of the N
stations are sending high priority data even though, in most cases, this document
at this time. For now, we propose only some issues
will mean that such delays are significantly overestimated.

Assuming that Controlled Load flows use a protocol
should consider:
* need to tackle problem of reservation request crossing traffic class lower than that
used by GS, no upper-bound on a access latency can be provided for CL
flows.  However, CL flows will receive better service than best effort
flows.

Note that, on many existing shared
   medium ("collisions"): this needs some form token rings, bridges will transmit
frames using an Access Priority (see section 3.3) value 4 irrespective
of tie- breaker.
* failed reservation retry policy: may be a bad idea to retry but we
   have to specify behaviour.
* one simple approach might be to avoid the election user_priority carried in the frame control field of any "master"
   bandwidth arbiter on a segment: if we were to assume an optimistic
   approach the frame.
Therefore, existing bridges would need to reservations with later "veto" power by subsequent
   switches be reconfigured or receivers then a large degree of complexity might modified
before the above access time bounds can actually be avoided.
* used.

12. Signaling protocol

The mechanisms described in this document make use of a signaling
protocol needs to be able for devices to notify failure of communicate their admission control back to client or back to previous switch hop.

12. Shared media

The astute reader will have noticed that we have not mentioned requests
across the
difficulty of dealing with allocation on network: the service definitions to be provided by such a single shared CSMA/CD
segment: there
protocol are a number of reasons described below. The candidate IETF protocol for this.

Firstly, we do not believe this

purpose is a truly solvable problem: it would
seem to require a new MAC protocol. Those who are interested called "Subnet Bandwidth Manager" and is described in solving
this problem per se should probably [10].

In all these cases, appropriate delete/cleanup mechanisms will also have
to be provided for when sessions are torn down. All interactions are
assumed to provide read as well as write capabilities.

12.1 Client service definitions

The following the BLAM developments
in 802.3 but we would be suspicious of the interoperability
characteristics of a series of new software MACs running above the
traditional 802.3 MAC.

Secondly, we interfaces are not convinced that it identified from Figures 2 and 3:

SBM <-> Address mapping

 This is really a simple lookup function which may cause ARP protocol
interactions, may be just a lookup of an interesting problem.
While not everyone existing ARP cache entry or may
be an algorithmic mapping. The layer-2 addresses are needed by SBM for
inclusion in its signaling messages to/from switches which avoids the world is buying desktop
switches today and
there will be end stations living on repeated segments for some time having to
come, perform the number mapping and, hence, have knowledge of switches is going up and
layer-3 information for the number of stations on
repeated segments is going down. complete subnet:

     l2_addr = map_address( ip_addr )

SBM <-> Session/802 header

This trend is proceeding to for notifying the point
that we may be happy transmit path of how to associate
user_priority values with a solution which assumes that any network
conversation requiring resource reservations the traffic of each outgoing session: the
transmit path will take place through at
least one switch (be it layer-2 or layer-3). Put another way, provide the
easiest QoS upgrade to user_priority value when it requests a layer-2 network
MAC-layer transmit operation for each packet (user_priority is to install segment
switching: only when has been done one of
the parameters defined by the IEEE 802 service model):

     bind_802_header( sessionid, user_priority )

SBM <-> Classifier/Scheduler

This is it worthwhile to investigate more
complex solutions involving admission control.

Thirdly, in for notifying transmit classifier/scheduler of additional
layer-2 information associated with scheduling the core transmission of a
session's packets (may be unused in some cases):

     bind_l2sessioninfo( sessionid, l2_header, traffic_class )

SBM <-> Local Admission Control

For applying local admission control for a session e.g. is there enough
transmit bandwidth still uncommitted for this potential new session? Are
there sufficient receive buffers? This should commit the network (as opposed necessary
resources if OK: it will be necessary to at release these resources if a
later stage of the edges), there
does not seem to be enough economic benefit session setup process fails.

     status = admit_l2txsession( Tspec, flowspec )
     status = admit_l2rxsession( Rspec, flowspec )

SBM <-> RSVP - this is outlined above in section 8.2 and fully described
in [10].

12.2 Switch service definitions

The following interfaces are identified from Figure 4:

SBM <-> Classifier

This is for repeated segment
solutions as opposed notifying receive classifier of how to switched solutions. While repeated solutions
*may* be 50% cheaper, their cost impact on match up incoming
layer-2 information with the entire network associated traffic class: it may in some
cases consist of a set of read-only default mappings:

     bind_l2classifierinfo( l2_header, traffic_class )

SBM <-> Queue and Packet Scheduler

This is
amortised across all for notifying transmit scheduler of the edge ports. There additional layer-2
information associated with a given traffic class (it may be special
circumstances unused in
some cases):

     bind_l2schedulerinfo( l2_header, traffic_class )

SBM <-> Local Admission Control

 As for host above.

SBM <-> Traffic Class Map and Police

 Optional configuration of any layer-2 policing function and/or
user_priority remapping that might be implemented on input to a switch:

     bind_l2classmapping( in_user_priority, remap_user_priority )
     bind_l2policing( l2_header, traffic_characteristics )

SBM <-> Filtering Database

SBM propagation rules need access to the future (e.g. Gigabit buffered repeaters) but these
have differing characteristics layer-2 forwarding database to existing CSMA/CD repeaters anyway.
determine where to forward SBM messages (analogous to RSRR interface in
L3 RSVP):

     output_portlist = lookup_l2dest( l2_addr )

13. Compatibility and Interoperability with existing equipment

Layer-2-only "standard" 802.1p switches will have to work together with
routers and layer-3 switches. Wide deployment of such 802.1p switches is
envisaged, in a number of roles in the network. "Desktop switches" will
provide dedicated 10/100 Mbps links to end stations at costs
comparable/compatible with NICs/adapter cards. Very high speed core
switches may act as central campus switching points for layer 3 devices.
Real network deployments provide a wide range of examples today. The
question is "what functionality beyond that of the basic 802.1D bridge
should such 802.1p switches provide?". In the abstract the answer is
"whatever they can do to broaden the applicability of the switching
solution while still being economically distinct from the layer 3
switches in their cost of acquisition, speed/bandwidth, cost of
ownership and administration". Broadening the applicability means both
addressing the needs of new traffic types and building larger switched
networks (or making larger portions of existing networks switched). Thus
one could imagine a network in which every device (along a network path)
was layer-3 capable/intrusive into the full data stream; or one in which
only the edge devices were pure layer-2; or one in which every alternate
device lacked layer-3 functionality; or most do - excluding some key
control points such as router firewalls, for example. Whatever the mix,
the solution has to interoperate with these layer-3 QoS-aware devices.

Of course, where intserv int-serv flows pass through equipment which is ignorant
of priority queuing and which places all packets through the same
queuing/overload-dropping path, it is obvious that some of the
characteristics of the flow get more difficult to support. Suitable
courses of action in the cases where sufficient bandwidth or buffering
is not available are of the form:

(a)  buy more (and bigger) routers
(b)  buy more capable switches
(c)  rearrange the network topology: 802.1Q VLANs [11] may help here.
(d)  buy more bandwidth: Gigabit Ethernet is nearly here. bandwidth

It would also be possible to pass more information between switches
about the capabilities of their neighbours and to route around non-
QoS-capable switches: such methods are for further study.

14. Epilogue Justification

An obvious comment is that this is all too complex, it's what RSVP is
doing already, why do we think we can do better by reinventing the
solution to this problem at layer-2?

The key is that we do not have to tackle the full problem space of RSVP:
there are a number of simple scenarios that cover a considerable
proportion of the real situations that occur: all we have to do here is
cover 99% of the territory at significantly lower cost and leave the
other applications to full RSVP running in strategically  positioned
high-function switches or routers. This will allow a significant
reduction in overall network cost (equipment and ownership). This
approach does mean that we have to discuss real life situations instead
of abstract topologies that "could happen".

Sometimes, for example, simple bandwidth configuration in a few switches
e.g. to avoid overloading particular trunk links, can be used to
overcome bottlenecks due to the network topology: if there are issues
with overloading end station "last hops", RSVP in the end stations would
exert the correct controls simply by examining local resources without
much tie-in to the layer-2 topology. In this case there has been no need
to resort to any form of complex topology computation and much
complexity has been avoided.

In the more general case, there remains work to be done. This will need
to be done against the background constraint that the changing of queue
service policies and the addition of extra functionality to support new
service disciplines will proceed at the rate of hardware product
development cycles and advance implementations of new algorithms may be
pursued reluctantly or without the necessary 20-20 20/20 foresight.

However, compared to the alternative of no traffic classes at all, there
is substantial benefit in even the simplest of approaches (e.g. 2-4
queues with straight priority), so there is significant reward for doing
something: wide acceptance of that "something" probably means that even
the simplest queue service disciplines will be provided for.

15. References

[1] ISO/IEC 10038, ANSI/IEEE Std 802.1D-1993 "MAC Bridges"

[2] "MAC Bridges - "Supplement to MAC Bridges: Traffic Classes Class Expediting and
       Dynamic Multicast Filtering
        Services in Bridged Local Area Networks", October 1996 Filtering",  May 1997, IEEE P802.1p/D4 P802.1p/D6

[3] "Integrated Services in the Internet Architecture: an Overview"

[4] "Resource Reservation Protocol (RSVP) - Version 1 Functional
       Specification" Internet Draft, November 1996
       <draft-ietf-rsvp-spec-14.ps
       <draft-ietf-rsvp-spec-16.[ps,txt]>

[5] "Carrier Sense Multiple Access with Collision Detection
       (CSMA/CD) Access Method and Physical Layer Specifications"
       ANSI/IEEE Std 802.3-1985.

[6] "Token-Ring Media Access Control"
         IEEE Method and Physical Layer Specifications"
       ANSI/IEEE Std 802.5 802.5-1995

[7] "A Framework for Providing Integrated Services Over Shared and
       Switched LAN Technologies", Internet Draft, November 1996
       <draft-ghanwani-framework-is-lan-01.txt
       <draft-ietf-issll-is802-framework-02>

[8] "Specification of the Controlled-Load Network Element Service",
       Internet Draft, August 1996,
       <draft-ietf-intserv-ctrl-load-svc-03.txt May 1997,
       <draft-ietf-intserv-ctrl-load-svc-05.txt>

[9] "Specification of Guaranteed Quality of Service",
       Internet Draft, August 1996,
       <draft-ietf-intserv-guaranteed-svc-06.txt February 1997,
       <draft-ietf-intserv-guaranteed-svc-07.txt>

[10] "SBM (Subnet Bandwidth Manager): A Proposal for Admission
       <draft-yavatkar-sbm-ethernet-04>

[11] "Draft Standard for Virtual Bridged Local Area Networks",
        May 1997, IEEE P802.1Q/D6

[12] "General Characterization Parameters for Integrated
        Service Network Elements", Internet Draft, November 1996
        <draft-ietf-intserv-charac-02.txt>

[13] "A Standard for the Transmission of IP Datagrams over IEEE
        802 Networks", RFC 1042, February 1988

[14] "The Use of Priorities on Token-Ring Networks for Multimedia
        Traffic", C. Bisdikian, B. V. Patel, F. Schaffa and M.
        Willebeek-LeMair, IEEE Network, Nov/Dec 1995.

16. Security Considerations

Security issues

There are not addressed no known security issues over and above those inherent in the
Integrated Services architecture and the network technologies referenced
by this memo. document.

17. Acknowledgments

This document draws heavily on the work of the ISSLL WG of the IETF and
the IEEE P802.1 Interworking Task Group. In particular, it includes
previous work on Token-Ring by Anoop Ghanwani, Wayne Pace and Vijay
Srinivasan.

18. Authors' addresses

Mick Seaman
3Com Corp.
5400 Bayfront Plaza
Santa Clara CA 95052-8145
USA
+1 (408) 764 5000
mick_seaman@3com.com

Andrew Smith
Extreme Networks
1601 S De Anza Blvd. #220
10460 Bandley Drive
Cupertino CA 95014
USA
+1 (408) 342 0999 863 2821
andrew@extremenetworks.com

Eric Crawley
Bay
Gigapacket Networks
3 Federal St.
Billerica
25 Porter Rd.
Littleton MA 01821 01460
USA
+1 (508) 670 8888
esc@baynetworks.com 486 0665
esc@gigapacket.com