Intent Driven Networks: Challenges and Enablers
•
4 j'aime•1,087 vues
Keynote speech at IEEE flagship conference on Integrated Network Management. Developing a comprehensive vision and understanding of Intent driven networks at Nokia Bell Labs.
Recommandé
Recommandé
Contenu connexe
Tendances
Tendances (20)
Similaire à Intent Driven Networks: Challenges and Enablers
Similaire à Intent Driven Networks: Challenges and Enablers (20)
Dernier
Dernier (20)
Intent Driven Networks: Challenges and Enablers
- 1. 1 © Nokia 2015 Intent-Driven Networks Challenges and Enablers Public IFIP/IEEE International Symposium on Integrated Network Management Lisbon // Portugal // 8-12 May 2017 Laurent Ciavaglia
- 2. 2 © Nokia 2015 1 Why Intent driven networking is important Agenda Public 2 What Definitions and activities 3 How Challenges and enablers
- 3. 3 © Nokia 2015 1 Why Intent driven networking is important Agenda Public 2 What Definitions and activities 3 How Challenges and enablers
- 4. 4 © Nokia 2015 Our industry challenge is network complexity Public
- 5. 5 © Nokia 2015 • is inherent to network as a system of (distributed) systems • will not disappear • will grow… The more we enable and expose new data and behaviors, the more we create complexity Complexity Public
- 6. 6 © Nokia 2015 • machine learning and reasoning have their roots in complexity • machine learning and reasoning are key elements of the solution Complexity is good Public
- 7. 7 © Nokia 2015 • think autonomics as a mean to hide/reduce complexity of network (micro-)management and related operations • think agent based modeling • think distributed problem solving • think abstractions • etc. Complexity can be tamed Public
- 8. 8 © Nokia 2015 The problem… We choose the wrong bottle of ketchup ! Public
- 9. 9 © Nokia 2015 Public
- 10. 10 © Nokia 2015 Networks need a radical shift in usability Public
- 11. 11 © Nokia 2015 Recognize intentions in any form of data Enabling a shift in usability Public
- 12. 12 © Nokia 2015 Intelligence drives automation and adaptation Enabling a shift in usability Public
- 13. 13 © Nokia 2015 Enabling a shift in usability Public Accountable for decisions and efficiency
- 14. 14 © Nokia 2015 The promises of intent-driven networks Public
- 15. 15 © Nokia 2015 • 1 intent = 1000’s command lines • alleviate the (imperative) policy explosion problem • lower personal training Create time! Performance gain Public
- 16. 16 © Nokia 2015 • simpler (?) policy conflict detection and resolution • greater flexibility in system response More challenging to identify (and quantify) but respectively more important Functionality gain Public
- 17. 17 © Nokia 2015 1 Why Intent driven networking is important Agenda Public 2 What Definitions and activities 3 How Challenges and enablers
- 18. 18 © Nokia 2015 Intent = what not how Public
- 19. 19 © Nokia 2015 Networks configuring and adapting autonomously to the user or operator intentions (i.e., a desired state or behavior) without the need to specify every technical detail of the process and operations to achieve it (i.e., the "machines" will figure out on their own how to realize the user goal). Intent-driven networks Public
- 20. 20 © Nokia 2015 Intent as an envelope of utility function Public ConstraintsIntents User Operator Network Capabilities fn ((max(utilityuser), min(costoperator))
- 21. 21 © Nokia 2015 Intent as a declarative policy Public Human Machine Imperative Declarative
- 22. 22 © Nokia 2015 Event – Condition – Action (ECA) Explicit programming of state rationality is compiled into the policy ! Pros: can be simple, system knows exactly what to do Cons: explosion of policies #, conflict detection and resolution can be very difficult, difficult to read, complex to write Imperative policy Public from J. Strassner
- 23. 23 © Nokia 2015 Express what should be done, not how to do it Specifies criteria for choosing a set of states rationality is generated by optimizer/planner Pros: more abstract, potentially more flexible, fewer, easier to write and comprehend Cons: requires sophisticated translation and optimization modules Declarative policy Public from J. Strassner
- 24. 24 © Nokia 2015 Intent Driven Networks Ingredients Public
- 25. 25 © Nokia 2015 Ingredients Public Intent Processing Engine Cognitive / Autonomic functions Virtualized and Programmable Infrastructure
- 26. 26 © Nokia 2015 Autonomic functions • intermediate abstraction points in the policy continuum • level(s) of autonomy to understand and react on intents self-adaptation and self-organization properties Ingredients mix Public
- 27. 27 © Nokia 2015 Software infrastructure • 2 root classes of actions: install, (re-configure) • if pre-determined capabilities discovery and configuration (by autonomic functions) • if virtual and programmable function placement, function composition Ingredients mix Public
- 28. 28 © Nokia 2015 You need more than good ingredients to make a good cake Public
- 29. 29 © Nokia 2015 We are now writing the cook book for intent driven network Public
- 30. 30 © Nokia 2015 The intent jungle Public
- 31. 31 © Nokia 2015 Huawei HPe Nokia Telefonica Orange Juniper Auvik NEMO The intent jungle Public SUPA Google ZTN MEF OpenStack Congress GBP ENI NMRG ONF Boulder Aspen ONOS Intent framework ANIMA Open Daylight NIC GBP APSTRA AOS Cisco Fortinet FortiOS
- 32. 32 © Nokia 2015 A reference document: Intent NBI – Definition and Principles https://www.opennetworking.org/images/stories/downloads/sdn- resources/technical-reports/TR-523_Intent_Definition_Principles.pdf Open Networking Foundation (ONF) Public
- 33. 33 © Nokia 2015 Open Networking Foundation (ONF) Public
- 34. 34 © Nokia 2015 Standardizing a reference framework and protocols for autonomic networks. IETF ANIMA working group Public https://datatracker.ietf.org/wg/anima/documents/
- 35. 35 © Nokia 2015 An abstract, declarative, high-level policy used to operate an autonomic domain (as per draft-ietf-anima-reference-model-04 and RFC7575) Intent lifecycle One Autonomic Network = Multiple Intents One Intent = Multiple Outputs Network operators/administrators writes Intents Autonomic Functions define what Intents they understand IETF ANIMA working group Public
- 36. 36 © Nokia 2015 Examples of intents (not ANIMA specific) -Do the right thing -Freeze network enrollment -Arrange VM guest distribution so that (CPU) utilization is < 70% -Assign prefixes to RAN nodes -Protect premium users traffic -Maximize energy savings IETF ANIMA working group Public
- 37. 37 © Nokia 2015 A proper analysis would require a complete survey of all these initiatives and contributions. Volunteers? Highly fragmented space, common denominator very small Most preeminent work on “Intent for SDN” addressing essentially a pure connectivity need Interestingly, “Intent for NFV” has a quite low activity threshold Scientific literature is scarce or spread over multiple domains / disciplines Observations Public
- 38. 38 © Nokia 2015 1 Why Intent driven networking is important Agenda Public 2 What Definitions and activities 3 How Challenges and enablers
- 39. 39 © Nokia 2015 Challenge 1 Public Intent generation and validation Sensori input User input and data Operator input and data Shared goals Patterns Ability to learn and reason on intent sets
- 40. 40 © Nokia 2015 Natural Language Processing (NLP) Ontologies and semantic analysis Lazy learning Knowledge representation and building: • Languages, templates, models… • graph databases, data dependencies… (some) Research directions Public
- 41. 41 © Nokia 2015 Challenge 2 Public Formulate and solve problem Operational constraints Shared goals Patterns Decisions Ability to optimize and re-optimize decisions
- 42. 42 © Nokia 2015 Automatization of the generation of resolution methods (by decomposition) Learning on choice of resolution method based on theoretical gain (some) Research directions Public
- 43. 43 © Nokia 2015 Challenge 3 Public Program generation, specification and verification Execution constraints Decisions Instructions Ability to generate and verify programs
- 44. 44 © Nokia 2015 Automatic program generation Program distribution towards agents and collective decision process on resolution approach considering local/global variables and constraints (some) Research directions Public
- 45. 45 © Nokia 2015 Challenge 4 Public Control points and run-time assessment mechanisms
- 46. 46 © Nokia 2015 Quality of Intent (QoI) evaluation framework Self-evaluation, self-testing mechanisms Increased role for telemetry and analytics (some) Research directions Public
- 47. 47 © Nokia 2015 Intent checking, normalization Intent recommendation, learning, optimization Intent extraction out of CLIs Multiple sources, formats, time and space of intents More or less explicit / implicit formulation of intents and many other interesting things Public
- 48. 48 © Nokia 2015 Scale the problem by one order of magnitude …IoT… Public
- 49. 49 © Nokia 2015 Conclusion Public