ANIMA L. Ciavaglia
Internet-Draft P. Peloso
Intended status: Standards Track Alcatel-Lucent
Expires: January 7, 2016 July 6, 2015

Autonomic Functions Coordination
draft-ciavaglia-anima-coordination-00.txt

Abstract

This document describes a management solution capable of avoiding conflicts between autonomic functions. The objective of such a solution is to avoid network instabilities, by insuring that the autonomic functions pursuing different goals will cooperate instead of antagonize each other. This document provides both requirements and specifications for such a solution.

Requirements Language

The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this document are to be interpreted as described in RFC 2119 [RFC2119].

Status of This Memo

This Internet-Draft is submitted to IETF 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 http://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 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 January 7, 2016.

Copyright Notice

Copyright (c) 2015 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 (http://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.

This document may not be modified, and derivative works of it may not be created, except to format it for publication as an RFC or to translate it into languages other than English.


Table of Contents

1. Introduction

The document Autonomic Networking: Definitions and Design Goals [RFC7575] explains the fundamental concepts behind Autonomic Networking, and defines the relevant terms in this space. The central concepts are Autonomic Nodes and Autonomic Functions.

An Autonomic Function is characterized by its implementing a closed control-loop, which we can summarize as successively: [I-D.behringer-anima-reference-model] details more the control loops).

  1. Gathers metrics monitored by network equipments (that could be Autonomic Nodes, but not limited to)
  2. Determines/computes new actions out of these inputs plus possibly some of the additional elements: e.g. contextual inputs, provided intents and gathered experience,
  3. Set the computed parameters values (from the previous actions) inside the appropriate network equipments,
  4. These new parameters values influence the network behavior, such that the metrics gathered by the autonomic function will evolve,

(Section 7.5 of

The Autonomic Functions are normally designed to stabilize (converge), at least when the network conditions are themselves stable. However, conflicting interactions among Autonomic Functions can create instabilities even when the network conditions have not varied.

The document A Reference Model for Autonomic Networking [I-D.behringer-anima-reference-model] describes the reference model of autonomic networks, by describing the architecture and enumerating fundamental blocks (either infrastructure pieces or enabling functionalities). One of these functionalities pertains to the concomitant execution of multiple autonomic functions in a safe way (i.e. avoiding conflicts between these different autonomic loops). Section 8 of [I-D.behringer-anima-reference-model] (Coordination between Autonomic Functions) provides a brief introduction to this functionality.

This document tackles this topic by successively:

  1. Explaining why such a functionality is needed,
  2. Detailing which objectives such a functionality should reach,
  3. Sketching a simple behavior of this function,
  4. Providing requirements on autonomic functions (a tentative list in this document version),
  5. Providing some specifications items (in this preliminary version, while future versions would provide specifications),

2. Problem Statement

The need to coordinate the joint behavior of autonomic functions arises from the need to cope with conflicting situations and to provide the operator with the ability to steer autonomic network performance to a given (intended) operational point.

Several interaction types exist among autonomic functions such as cooperation, dependency, or conflict (and possibly others [TBD]).

Cooperation happens when an autonomic function can improve the behavior or performance of another autonomic function, such as a traffic forecasting function used by a traffic allocation function.

Dependency happens when an autonomic function cannot work without another one being present or accessible in the autonomic network.

Conflicts among autonomic functions emerges from direct and indirect interactions. A metric value conflict is a conflict where one metric is influenced by parameters of different autonomic functions. A parameter value conflict is a conflict where one parameter is modified by different autonomic functions. A simple example of conflicting interaction between autonomic functions is the oscillations caused by an energy-saving function (which switches-off interfaces to reduce power consumption) and a load-balancing function (which switches-on interfaces to reduce link load).

Solving the coordination problem beyond one-by-one cases can rapidly become intractable if one considers networks composed of tens, hundreds or thousands of simultaneously interacting functions. Specifying a common functional block on coordination is a first step to address the problem in a systemic way.

3. Guiding principles

A coordination function appears as an essential component of the ANIMA reference model in order to achieve better control on the performance, stability and convergence of autonomic networks.

As guiding principles, the ANIMA coordination function should:

4. Initial sketch of a Coordination Function

For the sake of the following sections, this section is providing a rough description of the functioning of a coordination function, and how it organizes itself along the network time.

4.1. Preliminary assumptions

Autonomic functions do exist in different states corresponding to different steps in their life-cycle. The description of some of these steps is better understood by referring to the High level view of an Autonomic Network which is depicted in Figure 1 of [I-D.behringer-anima-reference-model], which Figure is copied below:

						
+- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - +
:            :       Autonomic Function 1        :                 :
: ASA 1      :      ASA 1      :      ASA 1      :          ASA 1  :
+- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - +
             :                 :                 :
             :   +- - - - - - - - - - - - - - +  :
             :   :   Autonomic Function 2     :  :
             :   :  ASA 2      :      ASA 2   :  :
             :   +- - - - - - - - - - - - - - +  :
             :                 :                 :
+- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - +
:                Autonomic Networking Infrastructure               :
+- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - +
+--------+   :    +--------+   :    +--------+   :        +--------+
| Node 1 |--------| Node 2 |--------| Node 3 |----...-----| Node n |
+--------+   :    +--------+   :    +--------+   :        +--------+

					

Figure 1: High level view of an Autonomic Network

Undeployed -
In this state, the Autonomic Function is a mere piece of software, which may not even be copied on any node, but which may well be the code of the Autonomic Service Agents (ASA) corresponding to this Autonomic Function.
Instantiated/Deployed -
In this state the Autonomic Function is deployed, which means the ASA are available in the Nodes and gathered together into an Autonomic Function. In this state the autonomic function is bind to a scope which is the part of the network on which the autonomic function is meant to perform its duties. As a first approximation, the scope matches the Nodes which receive instructions from one of the ASA gathered in the Autonomic Function.
Running -
In this state, the autonomic function is deployed and is executing its closed control loop, hence acting on network, by modifying Nodes parameters.

The above list of states is not meant to be exhaustive, and would be better expanded in a document dedicated to Autonomic Functions, nevertheless the distinctions between the three above states are unavoidable.

4.2. Algorithms for coordination

This sub-section does not intend to specify algorithms capable of achieving coordination between autonomic functions, but means to illustrate different ways of avoiding conflicts, we can briefly list the following families of algorithms:

Random token -
This algorithm is insuring that each autonomic function is executing its control-loop the one after the other, the sequence is following a random pattern.
Time separation -
This algorithm is insuring that each autonomic function is executing its control-loop at different rates, e.g. for 2 functions: one is running fast enough to have time to converge in between two iterations of the slower one (this algorithm requires proper settings with regards of the autonomic functions to coordinate).
Efficiency bids -
In this algorithm, each autonomic function predicts which improvement its executing of its control-loop would bring, hence the coordination algorithms, picks the autonomic function promising the "best" improvement, and grants it the right to execute.

4.3. Behavior of the coordination function

This function is expected to steer the network towards a better "operating" point, by avoiding/mitigating detrimental interactions between Autonomic Functions.

The first step of such a process is the identification of these interactions and their classification in order to determine which ones have to be handled (at least the problematic ones i.e. conflicting ones).

The second step is the gathering of the identified interactions in groups that can be handled together while insuring the proper behavior of the network. This step intends to avoid handling all the interactions in one raw, but possibly to split the whole problem in smaller pieces, easier to handle.

The third step is the instantiation of coordination mechanisms well suited to handle each groups of interactions previously identified. Hence these coordination mechanisms would control the autonomic functions in order to insure a network behavior matching the intents of the network operator.

4.3.1. Times of the identification of interactions between AF

As the coordination function handles autonomic functions, its working is related to the different states of autonomic functions, namely, build-time, deploy-time and run-time. Hence the coordination function also present a life-cycle consisting in these 3 different states , in which the coordination function behaves according to the following descriptions:

At build-time, a common description of the autonomic function attributes (metrics, parameters, actions, capabilities...) allows to construct a "static interaction map" from the a-priori knowledge that can be derived/inferred from the functions attributes relationship. The static interaction map can be used as a first element by the operator (or mechanism) to (pre-)define policies and priorities as coordination strategies to manage the a-priori conflicts identified.

At deploy-time, autonomic functions are deployed on the network (i.e. installed, configured, instantiated...) but are not yet active/acting on the network. At this stage, for each instance of the autonomic functions and on a per resource basis, an inventory of the metrics monitored, of the actions performed and their relationships can be realized, resulting in a "dynamic interaction map". The dynamic interaction map provides the basis to identify conflicts that will happen at run-time, categorize them and plan for the appropriate coordination strategies/mechanisms.

At run-time, conflicts happens and arbitration is driven by the coordination strategies and available mechanisms. This is also the stage where new dependencies can be observed and inferred, ultimately resulting in update of the dynamic interaction map and possible adaptation of the coordination strategies and mechanisms.

4.3.2. Times of the coordination of AF

TBC

4.4. Conclusions

Some of the previous elements impact directly the coordination function, some other imply capacities of external elements such as Autonomic Functions and the Autonomic Control Plane. This conclusion is briefly categorizing and summarizing those:

Requirements onto the AF -
a descriptor of metrics and parameters/actions: a generic way of describing the inputs and outputs of the closed control loop, in order to identify the interactions.
a life-cycle: to match the process of the coordination (shortly stated, interaction identification and then conflict solving).
a common command interface of the autonomic functions: for the coordination to control the pace at which an autonomic function executes its control loop.

Requirements onto the ACP -
a common representation of information and knowledge: a function used to build the interactions maps.

Requirements onto the Coordination Function -
interaction identification: a function in charge of identifying interactions
interaction grouping: a function coping with grouping the previously identified interactions, in bundles that can be managed independently (for scalability concerns)
supporting various coordination mechanisms: to have the freedom of picking the most appropriate one.
interaction solving: a function capable of handling an independent bundle of interactions by controling the implied autonomic functions according to the picked algorithms.

5. External Requirements

At this stage of the document, this section is merely providing a structure of its content.

In order to achieve the aforementioned goals (detailed in section Section 3) a Coordination Functional Block should bring the following features:

Guidelines, recommendations or BCPs can also be provided for the aspects pertaining to the coordination strategies and mechanisms.

The coordination function requires a certain set of elements to work properly such as the autonomic function descriptor and the interaction map(s).

5.1. Autonomic Function Descriptor (AFD)

The Autonomic Function Descriptor (AFD) should contain the following elements:

5.2. control/command interface of AF

The Autonomic Function could be guided in its executing of its control-loop by the coordination mechanism. The guidance could range from preventing the executing of the control loop, to letting run on its own. In the middle of the range, coordination mechanism could restrain the actions, halt the control-loop at a given state of the execution (before enforcement).

This section can be expanded in conjunction with Section 7.5 of [I-D.behringer-anima-reference-model] details more the control loops.

5.3. Interaction/Information Maps

The Autonomic Control Plane(ACP) should be able to provide a view of the interactions between metrics in order to build the interaction maps. This functionality is needed to identify that metrics are coupled. E.g. the capacity of a link and its load ratio are intimately coupled, and to identify interactions between autonomic function, having this knowledge may prove instrumental.

6. Specifications

The coordination function can be decomposed in the following sub-functions:

TBC.

7. Acknowledgements

This draft was written using the xml2rfc project.

This draft content builds upon work achieved during UniverSelf FP7 EU project.

The authors thank Dimitri Papadimitriou for his valuable comments.

8. IANA Considerations

This memo includes no request to IANA.

9. Security Considerations

TBC

10. References

10.1. Normative References

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

10.2. Informative References

[I-D.behringer-anima-reference-model] Behringer, M., Carpenter, B., Eckert, T., Ciavaglia, L., Liu, B., Jeff, J. and J. Strassner, "A Reference Model for Autonomic Networking", Internet-Draft draft-behringer-anima-reference-model-03, June 2015.
[RFC7575] Behringer, M., Pritikin, M., Bjarnason, S., Clemm, A., Carpenter, B., Jiang, S. and L. Ciavaglia, "Autonomic Networking: Definitions and Design Goals", RFC 7575, June 2015.

Authors' Addresses

Laurent Ciavaglia Alcatel-Lucent Villarceaux Nozay, 91460 FR EMail: laurent.ciavaglia@alcatel-lucent.com
Peloso Pierre Alcatel-Lucent Villarceaux Nozay, 91460 FR EMail: pierre.peloso@alcatel-lucent.com