Evolving Network Automation Techniques for RealTime Applications Evolving

Evolving Network Automation Techniques for Real-Time Applications Evolving Network Operations and Automation and the Move to 400 G and Beyond Craig Hill Distinguished Systems Engineer US Public Sector CTO Office CCIE #1628 @netwrkr 95 / crhill@cisco. com EIS 2019 May 1, 2019 - Atlanta

Evolving Network Automation Techniques for Real-Time Applications Agenda • What and Why Closed-Loop Automation • Introduction to Streaming Telemetry • Components of the Closed-Loop Architecture • The Move to 400 G Ethernet and Beyond • Resource References © 2019 Cisco and/or its affiliates. All rights reserved. Cisco Public 2

What is “Closed. Loop” Automation?

One-Way Automation and Programming • Automation tools greatly expedite configuration process • Variety of tools, both open source, and vendor specific • Configure devices and validate the operation • Challenge? Automation typically focuses “southbound” • Untapped use cases for two-way communications exist YANG Models NSO YDK Configuration - Create - Modify - Delete Network Element Virtual Router © 2019 Cisco and/or its affiliates. All rights reserved. Cisco Public 4

Telemetry Collection Has Been Slow and Hard… SNMP, syslog, CLI… • “PULL” model (vs. PUSH) • Too slow for real-time use cases • Data is unstructured text • Very network specific • Difficult to operationalize • Need faster, scalable, open, modeldriven management stacks to simplify how data is collected from network elements YANG Models YDK Telemetry Collection Telemetry - Pull model SNMP Syslog CLI Network Element (physical/virtual) © 2019 Cisco and/or its affiliates. All rights reserved. Cisco Public 7

“Scream Stream If You Wanna Go Faster” Telemetry: an automated communications process by which measurements and other data are collected at remote or inaccessible points and transmitted to receiving equipment for monitoring. https: //en. wikipedia. org/wiki/Telemetry Visibility and Analytics Edge router l Backbone router efu Us Data Center router sy a E Rel t s Fa iabl e As Aggregatio n router De mu ch te rm in as po s sib is t ic Branch le Peering router © 2019 Cisco and/or its affiliates. All rights reserved. Cisco Public 8

The “Push” Model for Telemetry Collection • • • There is a shift to a “PUSH” model for streaming telemetry YANG Models YDK Data is structured analytics ready for parsing/analysis (tool chains) Pipeline Offers “event” and “periodic” collection options • Various modes exist based on the app (dial-in, dial-out, event-driven) • Model-driven telemetry offers structured, independent transport, encapsulations and dictates what data is pushed Telemetry - Streaming - Push (vs. Pull) - Optimized Transport Network Element (physical/virtual) © 2019 Cisco and/or its affiliates. All rights reserved. Cisco Public 9

Two-Way Automation + Programming The Path to “Closed-Loop” Automation • • 3 rd party SIEM “Closed-Loop” leverages telemetry for intelligent automation based on events Events can trigger configurations (not just operators) • Telemetry gathered is read from some repo (e. g. , message bus) • Variety of tools, both open source and vendor specific • Offers new paradigm for Net. Ops, Sec. Ops, service deployments, and assuring service delivery YANG Models YDK Telemetry Collection Configuration - Create - Modify - Delete Telemetry - Streaming - Push (vs. Pull) - Optimized Transport Network Element (physical/virtual) © 2019 Cisco and/or its affiliates. All rights reserved. Cisco Public 10

Use Cases for the Community Closed-Loop Automation Using Model-Drive Telemetry (MDT) • Configure / Validate Examples – cloud peering, configurations that require dependencies to configure and operate • Security – leverage “closed-loop” security approaches, triggering actions based on telemetry (dynamic triggers vulnerabilities, DDo. S) • Qo. S – leverage Key Performance Indicators (KPIs) to trigger eventdriven push, then modifying classifiers for more optimized traffic assurance • ML/AI Consumption – Leverage the model-driven streaming telemetry framework to push network data fast, structured, so third-party SIEMs and other ML/AI process can rapidly leverage the data • Network Operations – Day 2+ - leverage self-learning data from the network elements, allowing proactive health monitor statistics of the network components (elements, links, applications), Cisco Cross. Work © 2019 Cisco and/or its affiliates. All rights reserved. Cisco Public 11

Components of the Closed Loop Infrastructure

OSI Model of Telemetry end-to-end Layer Data Collector Information collection for processing (e. g. , data monitoring, automation, analytics) Exporter Encoding and delivering data to the collector(s) destination(s) Producer Sending requested data in model format to the “Exporter” at defined intervals Data model Data store Raw data mapped to a model (YANG native, Open. Config, etc) Native (raw) data inside a router’s database © 2019 Cisco and/or its affiliates. All rights reserved. Cisco Public

The Model-Driven Manageability Stack Apps Controller Orchestrator SDK Model-Driven Telemetry Closed-loop automation Model-Driven Configuration Network Device App Model-Driven SDKs YANG Development Kit (YDK) Protocol NETCONF Encoding XML Transport Models App SSH g. RPC JSON TCP HOW ? GPB HTTP YANG Models (native, open) © 2018 Cisco and/or its affiliates. All rights reserved. Cisco Public HOW ? WHAT ?

Consume – Model-Driven Subscription on the Network Element (Where? How? What? Often? ) Example (IOS-XR): Dial-out Incremental Where(? ): Pipe. Line #View telemetry subscription in IOS-XR asbr 1# show run telemetry model-driven Where? How? What? How often? © 2019 Cisco and/or its affiliates. All rights reserved. Cisco Public 15

Where Am I Sending the Telemetry? ? A Basic Analytics Platform Architecture The first layer (Collection) – collects all the streams and transforms them from GPB/JSON into a format that will be supported by the layer above. Applications Storage Collection Pipeline The second layer (Storage) is usually a TSDB WHERE ? (time-series database). Takes the data from the layer below (collection) and keeps it together with timestamps. (think of telemetry as a big data solution) The third layer (Applications) is your “business logic” or tools that you will use with the data stored in a database. Data Center router Backbone router BNG Peering router Border router Each layer can be an open-sourced solution, a commercial app, or home-grown. © 2019 Cisco and/or its affiliates. All rights reserved. Cisco Public

Start Exploring Telemetry Architectures Now Getting Started With Open Source Tools Data Center router Backbone Router BNG Peering router Pipeline Border router How to build up the stack: https: //xrdocs. io/telemetry/tutorials/2018 -06 -04 -ios-xr-telemetry-collectionstack-intro © 2019 Cisco and/or its affiliates. All rights reserved. Cisco Public 22

© 2019 Cisco and/or its affiliates. All rights reserved. Cisco Public 24

Closed-Loop Automation Component Example Mission Applications and ML/AI Collection, Storage, Visibility REST YDK Cisco Crosswork • Cisco Network Service Orchestrator (NSO) • PCE Server g. RPC (YDK) Controller / Orchestrator / PCE Closed Loop Cycle Telemetry Collection and Storage g. RPC NETCONF PCEP © 2019 Cisco and/or its affiliates. All rights reserved. Cisco Public 25

© 2019 Cisco and/or its affiliates. All rights reserved. Cisco Public 26

Why Closed-Loop Automation Framework? Closedloop Automatio n • Focus on large-scale automation • Embrace multi-vendor network solutions • Focus on the mindset of the operator • Allow network operations to be as proactive as possible Intent Based Networking Improved Qo. E Accelerate Time to Identify Accelerate Time to Remediation Optimization and Remediation © 2019 Cisco and/or its affiliates. All rights reserved. Cisco Public

400 G Ethernet… and Beyond

High-speed Landscape Towards 400 G+… Solutions Optics & Optical Technology ASIC Industry standards and consortiums draw all of this together to create the necessary high-speed solutions components Market Trends © 2018 Cisco and/or its affiliates. All rights reserved. Cisco Public 30

Market Transition Barriers • Optics is a barrier to transition • Reduction of ASIC cost-per-bit is outpacing that of optics • Critical to 400 Gb. E transition that optics cost reductions are accelerated Cisco Public BRKOPT-2005

Key Areas Targeting 400 G Use Cases ( … 2020) Webscale Providers Hyper Scale-out fabrics, increased east/west traffic volume Federal / Enterprise Deployments Service / Transport Provider High performance I/O in Large DC’s Capacity growth Core/Edge/Agg High-speed core/edge, DC/Cloud interconnect (storage replication, HA) 100 G/ 400 G buffered spines/fabrics in SP DC & edge locations Ready for NFV / 5 G adoption cycle Inter DC Interconnect (hybrid) Increased need to support AI/ML applications/workloads at scale Lower per Gigabit Core/Edge mission growth capacity Transition from 10/40 G to 25/50/100 G server NICs Increase overall backbone capacity (mobile data, cloud) Silicon Innovation – Ultra Speed | Deep Buffers | Telemetry | 400 G Encryption © 2018 Cisco and/or its affiliates. All rights reserved. Cisco Public

Ethernet Roadmap 6 new speeds in 1 st 35 years of Ethernet 6 new speeds in a 2 year span • 2. 5 Gb. E – 2016 • 25 Gb. E – 2016 • 50 Gb. E – late 2018 • 200 Gb. E – 2017 • 400 Gb. E – 2017 • 25/50 GE Consortium • * MSAs • QSFP-DD, OSFP • 100 G Lambda * Multi-Source Agreements – new ones all the time. Not all get wide industry adoption © 2018 Cisco and/or its affiliates. All rights reserved. Cisco Public 35

800 G, but not 800 Gb. E ? ? Only a 2 x DC fabric speed step may not warrant all the churn that 800 Gb. E may bring • Leverage the return on investment on the 400 Gb. E building blocks Focus on dense 400 Gb. E next • Transition to 100 Gb/s Ser. Des • Transition to 7 nm PHY/DSP chips • Transition to 25. 6 Tb/s switches • 800 G bandwidth modules are likely (2 x 400 Gb. E, 8 x 100 Gb. E) 7 nm DSP 400 G Optics, ASIC and packaging technology is making it possible to consider moving the optics into the system or into the ASIC package. Cisco Public

Summary

Summary • Major shift towards streaming telemetry to collect data from network elements more quickly and efficiently • The model-driven manageability stack offers open source/standard transport/encoding methods with closed-loop automation methods for collecting structured data • Telemetry collection can be incremental, event driven, providing operators a streamlined framework for real-time, high-speed and scaled collection • Open source tools and telemetry stacks are available today • Shift beyond 400 G (to 800 G? ) has already started and may be in reduced time line compared to recent high-speed network transitions from past © 2019 Cisco and/or its affiliates. All rights reserved. Cisco Public 38

Resources and References

Useful Links/Tools, How to get started… • IOS-XR Collection Stack Intro: https: //xrdocs. io/telemetry/tutorials/2018 -06 -04 -ios-xr-telemetry-collection-stack-intro/ • IOS XR specific telemetry & programmability content: https: //xrdocs. io/ • Everything you need to know about pipeline: https: //xrdocs. io/telemetry/tutorials/2018 -03 -01 -everything-you-needto-know-about-pipeline/ • Reference collector: https: //github. com/cisco/bigmuddy-networktelemetry-pipeline BRKNMS-3537 © 2019 Cisco and/or its affiliates. All rights reserved. Cisco Public 40

Useful Links/Tools, How to get started… • IOS-XR Collection Stack Intro: https: //xrdocs. io/telemetry/tutorials/2018 -06 -04 -ios-xr-telemetry-collection-stack-intro/ • Advanced NETCONF Explorer: https: //github. com/cisco-ie/anx • Telemetry Data Mapper: https: //github. com/cisco-ie/tdm • YANG model catalog and search engine: https: //yangcatalog. org/ • YANG model collection on Github: https: //github. com/Yang. Models/yang • IOS XR specific telemetry & programmability content: https: //xrdocs. io/ • Reference collector: https: //github. com/cisco/bigmuddy-networktelemetry-pipeline BRKNMS-3537 © 2019 Cisco and/or its affiliates. All rights reserved. Cisco Public 41

Cisco Live Reference Sessions • BRKSPG-2303: Model-Driven Programmability for Cisco IOS XR • BRKSPG-2503: Advanced Topics in XR Telemetry • DEVWKS-2561: Hands-on Exploration of NETCONF and YANG • DEVWKS-2077: Using YANG Models and Telemetry for Closed-Loop Applications © 2019 Cisco and/or its affiliates. All rights reserved. Cisco Public 42

g. NMI Implementation in Cisco IOS XR • Based on g. NMI v 0. 4. 0 • Introduced in release 6. 5. 1 • Set and Get RPCs use JSON_IETF (RFC 7951) and ASCII (CLI) encoding • Subscribe RPC Paths must consider data aggregation points (no arbitrary paths) • No aliases • © 2019 Cisco and/or its affiliates. All rights reserved. Cisco Public

Getting Started with g. NMI in YDK 0. 8. 0 Native Virtual d. Cloud Install Vagrant Install Virtualbox Download ydk-py-samples Install Python Install YDK Download ydk-py-samples Install docker Download from Docker Hub YANG Development Kit Sandbox 3. 0 d. Cloud. cisco. com © 2018 Cisco and/or its affiliates. All rights reserved. Cisco Public

Streaming Telemetry Components (used in Demo) Pipeline: https: //github. com/cisco/bigmuddy-network-telemetry-pipeline Kafka: https: //github. com/apache/kafka Pipeline Telemetry Peering router © 2019 Cisco and/or its affiliates. All rights reserved. Cisco Public 45

400 G Resources • Mark Nowell Blog’s • • Cisco 400 G Blog’s • • https: //blogs. cisco. com/author/marknowell https: //blogs. cisco. com/tag/400 g Cisco 400 G Data Center Portfolio • https: //www. cisco. com/c/en/us/solutions/data-center/high-capacity-400 gdata-center-networking/index. html Cisco Public