[Docs] [txt|pdf|xml|html] [Tracker] [Email] [Nits]

Versions: 00 draft-pedro-nmrg-anticipated-adaptation

NFVRG                                             P. Martinez-Julia, Ed.
Internet-Draft                                                      NICT
Intended status: Informational                                2 20, 2018
Expires: August 24, 2018

Exploiting External Event Detectors to Anticipate Resource Requirements
             for the Elastic Adaptation of SDN/NFV Systems


   The adoption of SDN/NFV technologies by current computer and network
   system infrastructures is constantly increasing, becoming essential
   for the the particular case of edge/branch network systems.  The
   systems supported by these infrastructures require to be adapted to
   environment changes within a short period of time.  Thus, the
   complexity of new systems and the speed at which management and
   control operations must be performed go beyond human limits.  Thus,
   management systems must be automated.  However, in several situations
   current automation techniques are not enough to respond to
   requirement changes.  Here we propose to anticipate changes in the
   operation environments of SDN/NFV systems in response to external
   events and reflect it in the anticipation of the amount of resources
   required by those systems for their ulterior adaptaion.  The final
   objective is to avoid service degradation or disruption while keeping
   close-to-optimum resource allocation to reduce monetary and operative
   cost as much as possible.  Here we discuss how to achieve such
   capabilities by the integration of the Autonomic Resource Control
   Architecture (ARCA) to the management and operation (MANO) of NFV
   systems.  We showcase it by building a multi-domain SDN/NFV
   infrastructure based on OpenStack and deploying ARCA to adapt a
   virtual system based on the edge/branch network concept to the
   operational conditions of an emergency support service, which is
   rarely used but that cannot leave any user unattended.

Status of This Memo

   This Internet-Draft is submitted in full conformance with the
   provisions of BCP 78 and BCP 79.

   Internet-Drafts are working documents of the Internet Engineering
   Task Force (IETF).  Note that other groups may also distribute
   working documents as Internet-Drafts.  The list of current Internet-
   Drafts is at https://datatracker.ietf.org/drafts/current/.

   Internet-Drafts are draft documents valid for a maximum of six months
   and may be updated, replaced, or obsoleted by other documents at any

Martinez-Julia           Expires August 24, 2018                [Page 1]

Internet-Draft  Anticipated Adaptation of SDN/NFV Systems         2 2018

   time.  It is inappropriate to use Internet-Drafts as reference
   material or to cite them other than as "work in progress."

   This Internet-Draft will expire on August 24, 2018.

Copyright Notice

   Copyright (c) 2018 IETF Trust and the persons identified as the
   document authors.  All rights reserved.

   This document is subject to BCP 78 and the IETF Trust's Legal
   Provisions Relating to IETF Documents
   (https://trustee.ietf.org/license-info) in effect on the date of
   publication of this document.  Please review these documents
   carefully, as they describe your rights and restrictions with respect
   to this document.  Code Components extracted from this document must
   include Simplified BSD License text as described in Section 4.e of
   the Trust Legal Provisions and are provided without warranty as
   described in the Simplified BSD License.

Table of Contents

   1.  Introduction  . . . . . . . . . . . . . . . . . . . . . . . .   3
   2.  Terminology . . . . . . . . . . . . . . . . . . . . . . . . .   4
   3.  Background  . . . . . . . . . . . . . . . . . . . . . . . . .   4
     3.1.  Virtual Computer and Network Systems  . . . . . . . . . .   4
     3.2.  SDN and NFV . . . . . . . . . . . . . . . . . . . . . . .   5
     3.3.  Management and Control  . . . . . . . . . . . . . . . . .   5
     3.4.  The Autonomic Resource Control Architecture (ARCA)  . . .   6
   4.  External Event Detectors  . . . . . . . . . . . . . . . . . .   7
   5.  Anticipating Requirements . . . . . . . . . . . . . . . . . .   8
   6.  ARCA Integration With ETSI-NFV-MANO . . . . . . . . . . . . .   8
     6.1.  Functional Integration  . . . . . . . . . . . . . . . . .   9
     6.2.  Target Experiment and Scenario  . . . . . . . . . . . . .  11
     6.3.  OpenStack Platform  . . . . . . . . . . . . . . . . . . .  13
     6.4.  Initial Results . . . . . . . . . . . . . . . . . . . . .  14
   7.  Relation to Other IETF/IRTF Initiatives . . . . . . . . . . .  17
   8.  IANA Considerations . . . . . . . . . . . . . . . . . . . . .  17
   9.  Security Considerations . . . . . . . . . . . . . . . . . . .  17
   10. Acknowledgements  . . . . . . . . . . . . . . . . . . . . . .  17
   11. References  . . . . . . . . . . . . . . . . . . . . . . . . .  17
     11.1.  Normative References . . . . . . . . . . . . . . . . . .  18
     11.2.  Informative References . . . . . . . . . . . . . . . . .  18
   Author's Address  . . . . . . . . . . . . . . . . . . . . . . . .  19

Martinez-Julia           Expires August 24, 2018                [Page 2]

Internet-Draft  Anticipated Adaptation of SDN/NFV Systems         2 2018

1.  Introduction

   The incorporation of Software Defined Networking (SDN) and Network
   Function Virtualization (NFV) to current infrastructures to build
   virtual computer and network systems is constantly increasing.  The
   need to automate the management and control of such systems has
   motivated us to design the Autonomic Resource Control Architecture
   (ARCA), as presented in ICIN 2018 [ICIN-2018].  Automation
   requirements are enough justified by the increasing size and
   complexity of systems, which in turn are essential in the current
   digital world.  Moreover, the particular requirements and market
   benefits of network virtualization have been crystallized in the
   uprising of SDN/NFV infrastructures.  Nowadays they broad reception
   of the combined SDN/NFV technology supposes a huge leap towards the
   empowerment and homogenization of virtualization technologies.
   Therefore, we have modeled ARCA to fit within the reference
   architecture for management and orchestration of NFV elements, the
   Virtual Network Functions (VNFs).

   Behind the scenes, NFV is based on a highly distributed and network
   empowered version of the well-known Cloud infrastructures and
   platforms, also complemented by their centralized counterparts.  This
   takes to virtual networks the high degree of flexibility already
   found for computer systems.  It is highly desirable at the time NFV
   is being exploited by many organizations to build their private
   infrastructures, as well as by network service providers to build the
   services they later commercialize.  However, to actually exploit the
   potential monetary and operative cost reduction that is associated to
   such infrastructures, the amount of resources used by production
   services must be kept close to the optimum, so the physical resources
   are exploited as much as possible.

   The fast detection of changes in the requirements of the virtual
   systems deployed on the aforementioned SDN/NFV infrastructures, and
   the consequent adaptation of allocated resources to the new
   situations, becomes essential to actually exploit their cost and
   operative benefits, while also avoiding service unresponsiveness due
   to underlying resource overloading.  It is widely accepted that the
   size and complexity of systems and services makes it difficult for
   humans to accomplish such task within their objective time
   boundaries.  Therefore, they must be automated.  Luckily, the
   architecture and underlying platforms supporting the SDN/NFV
   technologies enable the required automation.  In fact, some solutions
   already exist to perform several batched or scripted tasks without
   human intervention.  However, those solutions still have high
   dependences on low-level human involvement.  This remarks the
   challenge found in control and management automation, which is
   continuously revised and enlarged.

Martinez-Julia           Expires August 24, 2018                [Page 3]

Internet-Draft  Anticipated Adaptation of SDN/NFV Systems         2 2018

   ARCA provides as a small step towards the resolution of the
   aforementioned problem.  It advances the State of the Art in
   automation of resource control and management by providing a
   supervised but autonomous mechanism that reduces the time required to
   perform corrective and/or adaptive changes in virtual computer and
   network systems from hours/minutes to seconds/milliseconds.
   Moreover, it is able to take advantage of the event notifications
   provided by external detectors to anticipate the amount of resources
   that the controlled SDN/NFV system will require in response to such
   event.  We propose to bring such benefit to the reference
   architecture promoted by ETSI for the management and orchestration of
   NFV services (see ETSI-NFV-MANO [ETSI-NFV-MANO]) by integrating ARCA
   as the Virtual Infrastructure Manager (VIM).  We showcase this
   proposal by discussing the evaluation results obtained by ARCA when
   runnion on a real and physical experimentation infrastructure based
   on OpenStack [OPENSTACK].  We thus justify the need to adapt the
   interfaces supported by the NFV-MANO to include real-world event
   detectors, which are external to the virtualization platform and
   virtual resources.

2.  Terminology

   The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT",
   document are to be interpreted as described in RFC 2119 [RFC2119].

3.  Background

3.1.  Virtual Computer and Network Systems

   The continuous search for efficiency and cost reduction to get the
   most optimum exploitation of available resources (e.g.  CPU power and
   electricity) has conducted current physical infrastructures to move
   towards virtualization infrastructures.  Also, this trend enables end
   systems to be centralized and/or distributed, so that they are
   deployed to best accomplish customer requirements in terms of
   resources and qualities.

   One of the key functional requirements imposed to computer and
   network virtualization is a high degree of flexibility and
   reliability.  Both qualities are subject to the underlying
   technologies but, while the latter has been always enforced to
   computer and network systems, flexibility is a relatively new
   requirement, which whould not have been impossed without the backing
   of virtualization and cloud technologies.

Martinez-Julia           Expires August 24, 2018                [Page 4]

Internet-Draft  Anticipated Adaptation of SDN/NFV Systems         2 2018

3.2.  SDN and NFV

   SDN and NFV are conceived to bring high degree of flexibility and
   conceptual centralization qualities to the network.  On the one hand,
   with SDN, the network can be programmed to implement a dynamic
   behavior that changes its topology and overall qualities.  Moreover,
   with NFV the functions that are typically provided by physical
   network equipment are now implemented as virtual appliances that can
   be deployed and linked together to provide customized network
   services.  SDN and NFV complements to each other to actually
   implement the network aspect of the aforementioned virtual computer
   and network systems.

   Although centralization can lead us to think on the single-point-of-
   failure concept, it is not the case for these technologoes.
   Conceptual centralization highly differs from centralized deployment.
   It brings all benefits from having a single point of decision but
   retaining the benefits from distributed systems.  For instance,
   control decisions in SDN can be centralized while the mechanisms that
   enforce such decisions into the network (SDN controllers) can be
   implemented as highly distributed systems.  The same approach can be
   applied to NFV.  Althoug network functions can be implemented in a
   central computing facility, they can take advantage of several
   replication and distribution techniques to achieve the properties of
   distributed systems.  Nevertheless, NFV also allows the deployment of
   functions on top of distributed systems, so they benefit from both
   distribution alternatives at the same time.

3.3.  Management and Control

   The introduction of virtualization into the computer and network
   system landscape has increased the complexity of both underlying and
   overlying systems.  On the one hand, virtualyzing underlying systems
   adds extra functions that must be managed propoerly to ensure the
   correct operation of the whole system, which not just encompasses
   underlying elements but also the virtual elements running on top of
   them.  Such functions are used to actually host the overlying virtual
   elements, so there is an indirect management operation that involves
   virtual systems.  Moreover, such complexities are inherited by final
   systems that get virtualized and deployed on top of those
   virtualization infrastructures.

   In parallel, virtual systems are empowered with additional, and
   widely exploited, functionality that must be managed correctly.  It
   is the case of the dynamic adaptation of virtual resources to the
   specific needs of their operation environments, or even the
   composition of distributed elements across heterogeneous underlying
   infrastructures, and probably providers.

Martinez-Julia           Expires August 24, 2018                [Page 5]

Internet-Draft  Anticipated Adaptation of SDN/NFV Systems         2 2018

   Taking both complex functions into account, either separately or
   jointly, makes clear that management requirements have greatly
   supassed the limits of humans, so automation has become essential to
   accomplish most common tasks.

3.4.  The Autonomic Resource Control Architecture (ARCA)

   As deeply discussed in ICIN 2018 [ICIN-2018], ARCA leverages the
   elastic adaptation of resources assigned to virtual computer and
   network systems by calculating or estimating their requirements from
   the analysis of load measurements and the detection of external
   events.  These events can be notified by physical elements (things,
   sensors) that detect changes on the environment, as well as software
   elements that analyze digital information, such as connectors to
   sources or analyzers of Big Data.  For instance, ARCA is able to
   consider the detection of an earthquake or a heavy rainfall to
   overcome the damages it can make to the controlled system.

   The policies that ARCA must enforce will be specified by
   administrators during the configuration of the control/management
   engine.  Then, ARCA continues running autonomously, with no more
   human involvement unless some parameter must be changed.  ARCA will
   adopt the required control and management operations to adapt the
   controlled system to the new situation or requirements.  The main
   goal of ARCA is thus to reduce the time required for resource
   adaptation from hours/minutes to seconds/milliseconds.  With the
   aforementioned statements, system administrators are able to specify
   the general operational boundaries in terms of lower and upper system
   load thresholds, as well as the minimum and maximum amount of
   resources that can be allocated to the controlled system to overcome
   any eventual situation, including the natural crossing of such

   ARCA functional goal is to run autonomously while the performance
   goal is to keep the resources assigned to the controlled resources as
   close as possible to the optimum (e.g. 5 % from the optimum) while
   avoiding service disruption as much as possible, keeping client
   request discard rate as low as possible (e.g. below 1 %).  To achieve
   both goals, ARCA relies on the Autonomic Computing (AC) paradigm, in
   the form of interconnected micro-services.  Therefore, ARCA includes
   the four main elements and activities defined by AC, incarnated as:

   Collector Is responsible of gathering and formatting the
             heterogeneous observations that will be used in the control

   Analyzer  Correlates the observations to each other in order to find
             the situation of the controlled system, especially the

Martinez-Julia           Expires August 24, 2018                [Page 6]

Internet-Draft  Anticipated Adaptation of SDN/NFV Systems         2 2018

             current load of the resources allocated to the system and
             the occurrence of an incident that can affect to the normal
             operation of the system, such as an earthquake that
             increases the traffic in an emergency-support system, which
             is the main target scenario studied in this paper.

   Decider   Determines the necessary actions to adjust the resources to
             the load of the controlled system.

   Enforcer  Requests the underlying and overlying infrastructure, such
             as OpenStack, to make the necessary changes to reflect the
             effects of the decided actions into the system.

   Being a micro-service architecture means that the different
   components are executed in parallel.  This allows such components to
   operate in two ways.  First, their operation can be dispatched by
   receiving a message from the previous service or an external service.
   Second, the services can be self-dispatched, so they can activate
   some action or send some message without being previously stimulated
   by any message.  The overall control process loops indefinitely and
   it is closed by checking that the expected effects of an action are
   actually taking place.  The coherence among the distributed services
   involved in the ARCA control process is ensured by enforcing a common
   semantic representation and ontology to the messages they exchange.

   ARCA semantics are built with the Resource Description Framework
   (RDF) and the Web Ontology Language (OWL), which are well known and
   widely used standards for the semantic representation and management
   of knowledge.  They provide the ability to represent new concepts
   without requiring to change the software, just plugin extensions to
   the ontology.  ARCA stores all its knowledge is stored in the
   Knowledge Base (KB), which is queried and kept up-to-date by the
   analyzer and decider micro-services.  It is implemented by Apache
   Jena Fuseki, which is a high-performance RDF data store that supports
   SPARQL through an HTTP/REST interface.  Being de-facto standards,
   both technologies enable ARCA to be easily integrated to
   virtualization platforms like OpenStack.

4.  External Event Detectors

   As mentioned above, current mechanisms used to achieve automated
   management and control rely only on the continuous monitoring of the
   resources they control or the underlying infrastructure that host
   them.  However, there are several other sources of information that
   can be exploited to make the systems more robust and efficient.  It
   is the case of the notifications that can be provided by physical or
   virtual elements or devices that are watching for specific events,
   hence called external event detectors.

Martinez-Julia           Expires August 24, 2018                [Page 7]

Internet-Draft  Anticipated Adaptation of SDN/NFV Systems         2 2018

   More specifically, although the notifications provided by these
   external event detectors are related to successes that occur outside
   the boundaries of the controlled system, such successes can affect
   the typical operation of controlled systems.  For instance, a heavy
   rainfall or snowfall can be detected and correlated to a huge
   increase in the amount of requests experienced by some emergency
   support service.

5.  Anticipating Requirements

   One of the main goals of the MANO mechanisms is to ensure the virtual
   computer and network system they manage meets the requirements
   established by their owners and administrators.  It is currently
   achieved by observing and analyzing the performance measurements
   obtained either by directly asking the resources forming the managed
   system of by asking the controllers of the underlying infrastructure
   that hosts such resources.  Thus, under changing or eventual
   situations, the managed system must be adapted to cope with the new
   requirements, incrasing the amount of resources assigned to it, or to
   make efficient use of available infrastructures, reducing the amount
   of resources assigned to it.

   However, the time required by the infrastructure to make effective
   the adaptations requested by the MANO mechanisms is longer than the
   time required by client requests to overload the system and make it
   discard further client requests.  This situation is generally
   undesired but particularly dangerous for some systems, such as the
   emergency support system mentioned above.  Therefore, in order to
   avoid the disruption of the service, the change in requirements must
   be anticipated to ensure that any adaptation has finished as soon as
   possible, preferably before the target system gets overloaded or

   Here we propose to integrate ARCA with NFV-MANO to take advantage of
   the notifications provided by the aforementioned external event
   detectors, by correlating them to the target amount of resources
   required by the managed system and enforcing the necessary
   adaptations beforehand, particularly before the system performance
   metrics have actually changed.

6.  ARCA Integration With ETSI-NFV-MANO

   In this section we describe how to fit ARCA on a general SDN/NFV
   underlying infrastructure and introduce a showcase experiment that
   demonstrates its operation on an OpenStack-based experimentation
   platform.  We first describe the integration of ARCA with the NFV-
   MANO reference architecture.  We contextualize the significance of
   this integration by describing an emergency support scenario that

Martinez-Julia           Expires August 24, 2018                [Page 8]

Internet-Draft  Anticipated Adaptation of SDN/NFV Systems         2 2018

   clearly benefits from it.  Then we proceed to detail the elements
   forming the OpenStack platform and finally we discuss some initial
   results obtained from them.

6.1.  Functional Integration

   The most important functional blocks of the NFV reference
   architecture promoted by ETSI (see ETSI-NFV-MANO [ETSI-NFV-MANO]) are
   the system support functions for operations and business (OSS/BSS),
   the element management (EM) and, obviously. the Virtual Network
   Functions (VNFs).  But these functions cannot exist without being
   instantiated on a specific infrastructure, the NFV infrastructure
   (NFVI), and all of them must be coordinated, orchestrated, and
   managed by the general NFV-MANO functions.

   Both the NFVI and the NFV-MANO elements are subdivided into several
   sub-components.  The NFVI has the underlying physical computing,
   storage, and network resources, which are sliced (seedraft-qiang-
   [draft-qiang-coms-netslicing-information-model-02] and draft-geng-
   coms-architecture-01 [draft-geng-coms-architecture-01]) and
   virtualized to conform the virtual computing, storage, and network
   resources that will host the VNFs.  In addition, the NFV-MANO is
   subdivided in the NFV Orchestrator (NFVO), the VNF manager (VNFM) and
   the Virtual Infrastructure Manager (VIM).  As their name indicates,
   all high-level elements and sub-components have their own and very
   specific objective in the NFV architecture.

   During the design of ARCA we enforced both operational and
   interfacing aspects to its main objectives.  From the operational
   point of view, ARCA processes observations to manage virtual
   resources, so it plays the role of the VIM mentioned above.
   Therefore, ARCA has been designed with appropriate interfaces to fit
   in the place of the VIM.  This way, ARCA provides the NFV reference
   architecture with the ability to react to external events to adapt
   virtual computer and network systems, even anticipating such
   adaptations as performed by ARCA itself.  However, some interfaces
   must be extended to fully enable ARCA to perform its work within the
   NFV architecture.

   Once ARCA is placed in the position of the VIM, it enhances the
   general NFV architecture with its autonomic management capabilities.
   In particular, it discharges some responsibilities from the VNFM and
   NFVO, so they can focus on their own business while the virtual
   resources are behaving as they expect (and request).  Moreover, ARCA
   improves the scalability and reliability of the managed system in
   case of disconnection from the orchestration layer due to some
   failure, network split, etc.  It is also achieved by the autonomic

Martinez-Julia           Expires August 24, 2018                [Page 9]

Internet-Draft  Anticipated Adaptation of SDN/NFV Systems         2 2018

   capabilities, which, as described above, are guided by the rules and
   policies specified by the administrators and, here, communicated to
   ARCA through the NFVO.  However, ARCA will not be limited to such
   operation so, more generally, it will accomplish the requirements
   established by the Virtual Network Operators (VNOs), which are the
   owners of the slice of virtual resources that is managed by a
   particular instance of NFV-MANO, and therefore ARCA.

   In addition to the operational functions, ARCA incorporates the
   necessary mechanisms to engage the interfaces that enable it to
   interact with other elements of the NFV-MANO reference architecture.
   More specifically, ARCA is bound to the Or-Vi (see ETSI-NFV-IFA-005
   [ETSI-NFV-IFA-005]) and the Nf-Vi (see ETSI-NFV-IFA-004
   [ETSI-NFV-IFA-004] and ETSI-NFV-IFA-019 [ETSI-NFV-IFA-019]).  The
   former is the point of attachment between the NFVO and the VIM while
   the latter is the point of attachment between the NFVI and the VIM.
   In our current design we decided to avoid the support for the point
   of attachment between the VNFM and the VIM, called Vi-Vnfm (see ETSI-
   NFV-IFA-006 [ETSI-NFV-IFA-006]).  We leave it for future evolutions
   of the proposed integration, that will be enabled by a possible
   solution that provides the functions of the VNFM required by ARCA.

   Through the Or-Vi, ARCA receives the instructions it will enforce to
   the virtual computer and network system it is controlling.  As
   mentioned above, these are specified in the form of rules and
   policies, which are in turn formatted as several statements and
   embedded into the Or-Vi messages.  In general, these will be high-
   level objectives, so ARCA will use its reasoning capabilities to
   translate them into more specific, low-level objectives.  For
   instance, the Or-Vi can specify some high-level statement to avoid
   CPU overloading and ARCA will use its innate and acquired knowledge
   to translate it to specific statements that specify which parameters
   it has to measure (CPU load from assigned servers) and which are
   their desired boundaries, in the form of high threshold and low
   threshold.  Moreover, the Or-Vi will be used by the NFVO to specify
   which actions can be used by ARCA to overcome the violation of the
   mentioned policies.

   All information flowing the Or-Vi interface is encoded and formatted
   by following a simple but highly extensible ontology and exploiting
   the aforementioned semantic formats.  This ensures that the
   interconnected system is able to evolve, including the replacement of
   components, updating (addition or removal) the supported concepts to
   understand new scenarios, and connecting external tools to further
   enhance the management process.  The only requirement to ensure this
   feature is to ensure that all elements support the mentioned ontology
   and semantic formats.  Although it is not a finished task, the
   development of semantic technologies allows the easy adaptation and

Martinez-Julia           Expires August 24, 2018               [Page 10]

Internet-Draft  Anticipated Adaptation of SDN/NFV Systems         2 2018

   translation of existing information formats, so it is expected that
   more and more software pieces become easily integrable with the ETSI-
   NFV-MANO [ETSI-NFV-MANO] architecture.

   In contrast to the Or-Vi interface, the Nf-Vi interface exposes more
   precise and low-level operations.  Although this makes it easier to
   be integrated to ARCA, it also makes it to be tied to specific
   implementations.  In other words, building a proxy that enforces the
   aforementioned ontology to different interface instances to
   homogenize them adds undesirable complexity.  Therefore, new
   components have been specifically developed for ARCA to be able to
   interact with different NFVIs.  Nevertheless, this specialization is
   limited to the collector and enforcer.  Moreover, it allows ARCA to
   have optimized low-level operations, with high improvement of the
   overall performance.  This is the case of the specific
   implementations of the collector and enforcer used with Mininet and
   Docker, which are used as underlying infrastructures in previous
   experiments described in ICIN 2017 [ICIN-2017].  Moreover, as
   discussed in the following section, this is also the case of the
   implementations of the collector and enforcer tied to OpenStack
   telemetry and compute interfaces, respectively.

   Although OpenStack still lacks some functionality regarding the
   construction of specific virtual networks, we use it as the NFVI
   functional block in the integrated approach.  Therefore, OpenStack is
   the provider of the underlying SDN/NFV infrastructure and we
   exploited its APIs and SDK to achieve the integration.  More
   specifically, in our showcase we use the APIs provided by Ceilometer,
   Gnocchi, and Compute services as well as the SDK provided for Python.
   All of them are gathered within the Nf-Vi interface.  Moreover, we
   have extended the Or-Vi interface to connect external elements, such
   as the physical or environmental event detectors and Big Data
   connectors, which is becoming a mandatory requirement of the current
   virtualization ecosystem and it conforms our main extension to the
   NFV architecture.

6.2.  Target Experiment and Scenario

   From the beginning of our work on the design of ARCA we are targeting
   real-world scenarios, so we get better suited requirements.  In
   particular we work with a scenario that represents an emergency
   support service that is hosted on a virtual computer and network
   system, which is in turn hosted on the distributed virtualization
   infrastructure of a medium-sized organization.  The objective is to
   clearly represent an application that requires high dynamicity and
   high degree of reliability.  The emergency support service
   accomplishes this by being barely used when there is no incident but
   also being heavily loaded when there is an incident.

Martinez-Julia           Expires August 24, 2018               [Page 11]

Internet-Draft  Anticipated Adaptation of SDN/NFV Systems         2 2018

   Both the underlying infrastructure and virtual network share the same
   topology.  They have four independent but interconnected network
   domains that form part of the same administrative domain
   (organization).  The first domain hosts the systems of the
   headquarters (HQ) of the owner organization, so the VNFs it hosts
   (servants) implement the emergency support service.  We defined them
   as ``servants'' because they are Virtual Machine (VM) instances that
   work together to provide a single service by means of backing the
   Load Balancer (LB) instances deployed in the separate domains.  The
   amount of resources (servants) assigned to the service will be
   adjusted by ARCA, attaching or detaching servants to meet the load
   boundaries specified by administrators.

   The other domains represent different buildings of the organization
   and will host the clients that access to the service when an incident
   occurs.  They also host the necessary LB instances, which are also
   VNFs that are controlled by ARCA to regulate the access of clients to
   servants.  All domains will have physical detectors to provide
   external information that can (and will) be correlated to the load of
   the controlled virtual computer and network system and thus will
   affect to the amount of servants assigned to it.  Although the
   underlying infrastructure, the servants, and the ARCA instance are
   the same as those those used in the real world, both clients and
   detectors will be emulated.  Anyway, this does not reduce the
   transferability of the results obtained from our experiments as it
   allows to expand the amount of clients beyond the limits of most
   physical infrastructures.

   Each underlying OpenStack domain will be able to host a maximum of
   100 clients, as they will be deployed on a low profile virtual
   machine (flavor in OpenStack).  In general, clients will be
   performing requests at a rate of one request every ten seconds, so
   there would be a maximum of 30 requests per second.  However, under
   the simulated incident, the clients will raise their load to reach a
   common maximum of 1200 requests per second.  This mimics the shape
   and size of a real medium-size organization of about 300 users that
   perform a maximum of four requests per second when they need some

   The topology of the underlying network is simplified by connecting
   the four domains to the same, high-performance switch.  However, the
   topology of the virtual network is built by using direct links
   between the HQ domain and the other three domains.  These are
   complemented by links between domains 2 and 3, and between domains 3
   and 4.  This way, the three domains have three paths to reach the HQ
   domain: a direct path with just one hop, and two indirect paths with
   two and three hops, respectively.

Martinez-Julia           Expires August 24, 2018               [Page 12]

Internet-Draft  Anticipated Adaptation of SDN/NFV Systems         2 2018

   During the execution of the experiment, the detectors notify the
   incident to the controller as soon as it happens.  However, although
   the clients are stimulated at the same time, there is some delay
   between the occurrence of the incident and the moment the network
   service receives the increase in the load.  One of the main targets
   of our experiment is to study such delay and take advantage of it to
   anticipate the amount of servants required by the system.  We discuss
   it below.

   In summary, this scenario highlights the main benefits of ARCA to
   play the role of VIM and interacting with the underlying OpenStack
   platform.  This means the advancement towards an efficient use of
   resources and thus reducing the CAPEX of the system.  Moreover, as
   the operation of the system is autonomic, the involvement of human
   administrators is reduced and, therefore, the OPEX is also reduced.

6.3.  OpenStack Platform

   The implementation of the scenario described above reflects the
   requirements of any edge/branch networking infrastructure, which are
   composed of several distributed micro-data-centers deployed on the
   wiring centers of the buildings and/or storeys.  We chose to use
   OpenStack to meet such requirements because it is being widely used
   in production infrastructures and the resulting infrastructure will
   have the necessary robustness to accomplish our objectives, at the
   time it reflects the typical underlying platform found in any SDN/NFV

   We have deployed four separate network domains, each one with its own
   OpenStack instantiation.  All domains are totally capable of running
   regular OpenStack workload, i.e. executing VMs and networks, but, as
   mentioned above, we designate the domain 1 to be the headquarters of
   the organization.  The different underlying networks required by this
   (quite complex) deployment are provided by several VLANs within a
   high-end L2 switch.  This switch represents the distributed network
   of the organization.  Four separated VLANs are used to isolate the
   traffic within each domain, by connecting an interface of OpenStack's
   controller and compute nodes.  These VLANs therefore form the
   distributed data plane.  Moreover, other VLAN is used to carry the
   control plane as well as the management plane, which are used by the
   NFV-MANO, and thus ARCA.  It is instantiated in the physical machine
   called ARCA Node, to exchange control and management operations in
   relation to the collector and enforcer defined in ARCA.  This VLAN is
   shared among all OpenStack domains to implement the global control of
   the virtualization environment pertaining to the organization.
   Finally, other VLAN is used by the infrastructure to interconnect the
   data planes of the separated domains and also to allow all elements

Martinez-Julia           Expires August 24, 2018               [Page 13]

Internet-Draft  Anticipated Adaptation of SDN/NFV Systems         2 2018

   of the infrastructure to access the Internet to perform software
   installation and updates.

   Installation of OpenStack is provided by the Red Hat OpenStack
   Platform, which is tightly dependent on the Linux operating system
   and closely related to the software developed by the OpenStack Open
   Source project.  It provides a comprehensive way to install the whole
   platform while being easily customized to meet our specific
   requirements, while it is also backed by operational quality support.

   The ARCA node is also based on Linux but, since it is not directly
   related to the OpenStack deployment, it is not based on the same
   distribution.  It is just configured to be able to access the control
   and management interfaces offered by OpenStack, and therefore it is
   connected to the VLAN that hosts the control and management planes.
   On this node we deploy the NFV-MANO components, including the micro-
   services that form an ARCA instance.

   In summary, we dedicate nine physical computers to the OpenStack
   deployment, all are Dell PowerEdge R610 with 2 x Xeon 5670 2.96 GHz
   (6 core / 12 thread) CPU, 48 GiB RAM, 6 x 146 GiB HD at 10 kRPM, and
   4 x 1 GE NIC.  Moreover, we dedicate an additional computer with the
   same specification to the ARCA Node.  We dedicate a less powerful
   computer to implement the physical router because it will not be
   involved in the general execution of OpenStack nor in the specific
   experiments carried out with it.  Finally, as detailed above, we
   dedicate a high-end physical switch, an HP ProCurve 1810G-24, to
   build the interconnection networks.

6.4.  Initial Results

   Using the platform described above we execute an initial but long-
   lasting experiment based on the target scenario introduced at the
   beginning of this section.  The objective of this experiment is
   twofold.  First, we aim to demonstrate how ARCA behaves in a real
   environment.  Second, we aim to stress the coupling points between
   ARCA and OpenStack, which will raise the limitations of the existing

   With such objectives in mind, we define a timeline that will be
   followed by both clients and external event detectors.  It forces the
   virtualized system to experience different situations, including
   incidents of many severities.  When an incident is found in the
   timeline, the detectors notify it to the ARCA-based VIM and the
   clients change their request rates, which will depend on the severity
   of the incident.  This behavior is widely discussed in ICIN 2018
   [ICIN-2018], remarking how users behave after occurring a disaster or
   another similar incident.

Martinez-Julia           Expires August 24, 2018               [Page 14]

Internet-Draft  Anticipated Adaptation of SDN/NFV Systems         2 2018

   The ARCA-based VIM will know the occurrence of the incident from two
   sources.  First, it will receive the notification from the event
   detectors.  Second, it will notice the change of the CPU load of the
   servants assigned to the target service.  In this situation, ARCA has
   different opportunities to overcome the possible overload (or
   underload) of the system.  We explore the anticipation approach
   deeply discussed in ICIN 2018 [ICIN-2018].  Its operation is enclosed
   in the analyzer and decider and it is based on an algorithm that is
   divided in two sub-algorithms.

   The first sub-algorithm reacts to the detection of the incident and
   ulterior correlation of its severity to the amount of servants
   required by the system.  This sub-algorithm hosts the regression of
   the learner, which is based on the SVM/SVR technique, and predicts
   the necessary resources from two features: the severity of the
   incident and the time elapsed from the moment it happened.  The
   resulting amount of servants is established as the minimum amount
   that the VIM can use.

   The second sub-algorithm is fed with the CPU load measurements of the
   servants assigned to the service, as reported by the OpenStack
   platform.  With this information it checks whether the system is
   within the operating parameters established by the NFVO.  If not, it
   adjusts the resources assigned to the system.  It also uses the
   minimum amount established by the other sub-algorithm as the basis
   for the assignation.  After every correction, this algorithm learns
   the behavior by adding new correlation vectors to the SVM/SVR

   When the experiment is running, the collector component of the ARCA-
   based VIM is attached to the Telemetry interface of OpenStack by
   using the SDK to access the measurement data generated by Ceilometer
   and stored by Gnocchi.  In addition, it is attached to the external
   event detectors in order to receive their notifications.  On the
   other hand, the enforcer component is attached to the Compute
   interface of OpenStack by also using its SDK to request the
   infrastructure to create, destroy, query, or change the status of a
   VM that hosts a servant of the controlled system.  Finally, the
   enforcer also updates the lists of servers used by the load balancers
   to distribute the clients among the available resources.

   During the execution of the experiment we make the ARCA-based VIM to
   report the severity of the last incident, if any, the time elapsed
   since it occurred, the amount of servants assigned to the controlled
   system, the minimum amount of servants to be assigned, as determined
   by the anticipation algorithm, and the average load of all servants.
   In this instance, the severities are spread between 0 (no incident)
   and 4 (strongest incident), the elapsed times are less than 35

Martinez-Julia           Expires August 24, 2018               [Page 15]

Internet-Draft  Anticipated Adaptation of SDN/NFV Systems         2 2018

   seconds, and the minimum server assignation (MSA) is below 10,
   although the hard maximum is 15.

   With such measurements we illustrate how the learned correlation of
   the three features (dimensions) mentioned above is achieved.  Thus,
   when there is no incident (severity = 0), the MSA is kept to the
   minimum.  In parallel, regardless of the severity level, the
   algorithm learned that there is no need to increase the MSA for the
   first 5 or 10 seconds.  This shows the behavior discussed in this
   paper, that there is a delay between the occurrence of an event and
   the actual need for updated amount of resources, and it forms one
   fundamental aspect of our research.

   By inspecting the results, we know that there is a burst of client
   demands that is centered (peak) around 15 seconds after the
   occurrence of an incident or any other change in the accounted
   severity.  We also know that the burst lasts longer for higher
   severities, and it fluctuates a bit for the highest severities.
   Finally, we can also notice that for the majority of severities, the
   increased MSA is no longer required after 25 seconds from the time
   the severity change was notified.

   All that information becomes part of the knowledge of ARCA and it is
   stored both by the internal structures of the SVM/SVR and, once
   represented semantically, in the semantic database that manages the
   knowledge base of ARCA.  Thus, it is used to predict any future
   behavior.  For instance, is an incident of severity 3 has occurred 10
   seconds ago, ARCA knows that it will need to set the MSA to 6
   servants.  In fact, this information has been used during the
   experiment, so we can also know the accuracy of the algorithm by
   comparing the anticipated MSA value with the required value (or even
   the best value).  However, the analysis of such information is left
   for the future.

   While preparing and executing the experiment we found several
   limitation intrinsic to the current OpenStack platform.  First,
   regardless of the CPU and memory resources assigned to the underlying
   controller nodes, the platform is unable to record and deliver
   performance measurements at a lower interval than every 10 seconds,
   so it is currently not suitable for real time operations, which is
   important for our long-term research objectives.  Moreover, we found
   that the time required by the infrastructure to create a server that
   hosts a somewhat heavy servant is around 10 seconds, which is too far
   from our targets.  Although these limitations can be improved in the
   future, they clearly justify that our anticipation approach is
   essential for the proper working of a virtual system and, thus, the
   integration of external information becomes mandatory for future

Martinez-Julia           Expires August 24, 2018               [Page 16]

Internet-Draft  Anticipated Adaptation of SDN/NFV Systems         2 2018

   system management technologies, especially considering the
   virtualization environments.

   Finally, we found it difficult for the required measurements to be
   pushed to external components, so we had to poll for them.
   Otherwise, some component of ARCA must be instantiated along the main
   OpenStack components and services so it has first-hand and prompt
   access to such features.  This way, ARCA could receive push
   notifications with the measurements, as it is for the external
   detectors.  This is a key aspect that affects the placement of the
   NFV-VIM, or some subpart of it, on the general architecture.
   Therefore, for future iterations of the NFV reference architecture,
   an integrated view between the VIM and the NFVI could be required to
   reflect the future reality.

7.  Relation to Other IETF/IRTF Initiatives


8.  IANA Considerations

   This memo includes no request to IANA.

9.  Security Considerations

   The major security concerns of the integration of external event
   detectors and ARCA to manage SDN/NFV systems is that the boundaries
   of the control and management planes are crossed to introduce
   information from outside.  Such communications must be highly and
   heavily secured since some malfunction or explicit attacks might
   compromise the integrity and execution of the controlled system.
   However, it is up to implementers to deploy the necessary
   countermeasures to avoid such situations.  From the design point of
   view, since all oprations are performed within the control and/or
   management planes, the security level of the current solution is
   inherited and thus determined by the security masures established by
   the systems conforming such planes.

10.  Acknowledgements


11.  References

Martinez-Julia           Expires August 24, 2018               [Page 17]

Internet-Draft  Anticipated Adaptation of SDN/NFV Systems         2 2018

11.1.  Normative References

              "Technology Independent Information Model for Network
              Slicing", 2018, <https://www.ietf.org/id/

              "Technology Independent Information Model for Network
              Slicing", 2018, <https://www.ietf.org/id/

   [RFC2119]  Bradner, S., "Key words for use in RFCs to Indicate
              Requirement Levels", BCP 14, RFC 2119,
              DOI 10.17487/RFC2119, March 1997,

11.2.  Informative References

              ETSI NFV GS NFV-IFA 004, "Network Functions Virtualisation
              (NFV); Acceleration Technologies; Management Aspects
              Specification", 2016.

              ETSI NFV GS NFV-IFA 005, "Network Functions Virtualisation
              (NFV); Management and Orchestration; Or-Vi reference point
              - Interface and Information Model Specification", 2016.

              ETSI NFV GS NFV-IFA 006, "Network Functions Virtualisation
              (NFV); Management and Orchestration; Vi-Vnfm reference
              point - Interface and Information Model Specification",

              ETSI NFV GS NFV-IFA 019, "Network Functions Virtualisation
              (NFV); Acceleration Technologies; Management Aspects
              Specification; Release 3", 2017.

              ETSI NFV GS NFV-MAN 001, "Network Functions Virtualisation
              (NFV); Management and Orchestration", 2014.

Martinez-Julia           Expires August 24, 2018               [Page 18]

Internet-Draft  Anticipated Adaptation of SDN/NFV Systems         2 2018

              P. Martinez-Julia, V. P. Kafle, and H. Harai, "Achieving
              the autonomic adaptation of resources in virtualized
              network environments, in Proceedings of the 20th ICIN
              Conference (Innovations in Clouds, Internet and Networks,
              ICIN 2017). Washington, DC, USA: IEEE, 2018, pp. 1--8",

              P. Martinez-Julia, V. P. Kafle, and H. Harai,
              "Anticipating minimum resources needed to avoid service
              disruption of emergency support systems, in Proceedings of
              the 21th ICIN Conference (Innovations in Clouds, Internet
              and Networks, ICIN 2018). Washington, DC, USA: IEEE, 2018,
              pp. 1--8", 2018.

              The OpenStack Project, "http://www.openstack.org/", 2018.

Author's Address

   Pedro Martinez-Julia (editor)
   4-2-1, Nukui-Kitamachi
   Koganei, Tokyo  184-8795

   Phone: +81 42 327 7293
   Email: pedro@nict.go.jp

Martinez-Julia           Expires August 24, 2018               [Page 19]

Html markup produced by rfcmarkup 1.129b, available from https://tools.ietf.org/tools/rfcmarkup/