Model-driven network automationAnees Shaikh, Josh Georgeon behalf of Google Network Operations

2016 DevOps Networking Forum

Zero-touch networking : extreme automation

2

● network operations fully automated based on operator-expressed intent

● introducing new technologies is transparent to the operator interface

● automated detection of unintended behavior and rollback

● automation complies with network policies for safety and security

Network automation has come a long way ...

3

per-device automationCLI scripts

expect/ssh

unstructuredtext

NMSCLI engine

sshRPC API

NMS

automation library

drivers, templates

ssh vendor API

NMS

automation framework

ssh vendor API

recipes,modules,...

NMS

… but still work to do

● automation frameworks are great, but mostly just type CLI commands faster○ no change to the fundamental issue of multiple incompatible interfaces for the same things

● significant amount of proprietary integration to write and maintain○ true even if vendors provide the “drivers”, modules, recipes, playbooks, etc.

● introducing new platforms is considerable amount of new work

● what about visibility and monitoring ?

○ SNMP still the state of the art despite scaling limitations, legacy implementations, rigid structure, poor support for discovery, ...

4

Model-based automation

5

● durable APIs for managing and monitoring the network

● management abstraction layer (insulate from lower level details)

● forward compatibility with new platforms and technologies

● establishes a contract between NMS and infrastructure

Toward a vendor-neutral, model-driven world

6

Per-vendor tools

EMS A

platform-specific tools, processes, skills

tools B

EMS C

vendor A

vendor B

vendor C

Common OSS

EMS A

proprietary integrations, common interfaces upstream

EMS C

operator / 3rd party NMS

vendor A

vendor B

vendor C

Common mgmt APIs

common mgmt model

common management API, no proprietary integrations, native support on all vendors

vendor A

vendor B

vendor C

NMS

A model-based configuration pipeline

7

configuration datavendor-neutral, validated

operators

intent API“drain peering link”

update topology modeltopology models

config models

configurationgeneration

multiple vendor devices

gRPC req

gRPC endpoint

Streaming telemetry -- moving forward from SNMP

● stream data continuously -- with incremental updates

● telemetry sent based on subscriptions

● observe network state through a time-series data stream

● device data follows a common model

● efficient, secure transport protocols (gRPC)

8gRPC endpoint

telemetrycollectors

upstream systems

time-series data stream

async eventreporting

requests for ad-hoc data

Aim to deprecate SNMP in our network by 2017

opstate models

opstate models

OpenConfig : user-defined models

9

● informal industry collaboration among network operators

● data models for configuration and operational state, code written in YANG

● organizational model: informal, structured like an open source project

● development priorities driven by operator requirements

● engagements with major equipment vendors to drive native implementations

● engagement with standards (IETF) and OSS (ODL, ONOS, goBGP, Quagga)

TeraStream

OpenConfig progress

10

Published OpenConfig models

● BGP*● routing policy*● locally generated routes*● interfaces, aggregates, IP, VLANs*● MPLS, RSVP / TE*● networking / forwarding instances, VRFs● RIB contents● terminal optics*● streaming telemetry configuration*● top-level device structure

Models adopted by IETF for standardization* native platform implementations available or in progress

Models in development / review

● system inventory / hardware● ACLs● line optics (EDFAs and ROADMs)● system management● IS-IS● QoS● tunnels / encapsulation● LLDP● FIB / LFIB

Additional publications

● Reference RPC specification● YANG authoring style guide● Guidelines for versioning models

Models structured for observing network behavior

11

+--rw interface* [name] +--rw config | +--rw type | +--rw mtu? | +--rw name? | +--rw description? | +--rw enabled? +--ro state | +--ro type | +--ro mtu? | +--ro name? | +--ro description? | +--ro enabled? | +--ro oper-status | +--ro last-change? | +--ro counters | +--ro in-octets? | +--ro in-unicast-pkts? | +--ro in-broadcast-pkts? | +--ro in-multicast-pkts? | +--ro in-discards? | +--ro in-errors? | +--ro in-unknown-protos? | +--ro out-octets?

opstate = applied configuration + derived stated (counters, etc.)

data models structured to ease programmatic access to related operational state

opstate has deterministic, explicit location in the model paths

common data model

NMSinte

nded

v. a

pplie

d co

nfig

NMS acts on streaming view of state

quickly determine discrepancy between intended and applied configuration

code to exploit this pattern in multiple operators

OSS stack for model-based programmatic configuration

OpenConfig models

pyang

pyangbind

python classbindings

vendor A

vendor B

vendor C

template-based translations

12

validated, vendor-neutral object representationbgp.global.as = 15169

bgp.neighbors.neighbor.add(neighbor_addr=124.25.2.1”)

...

interfaces.interface.add("eth0")

eth0 = src_ocif.interfaces.interface["eth0"]

eth0.config.enabled = True

eth0.ethernet.config.duplex_mode = "FULL"

eth0.ethernet.config.auto_negotiate = True

config DB

vendor A

vendor B

vendor C

pybindJSON.dumps(bgp.neighbors)pybindJSON.dumps(interfaces)

gRPC / RESTCONF

vendor neutral

data collector(fluentd)

OSS stack for model-based streaming telemetry

Platform support for streaming telemetry:

Cisco IOS-XRgithub.com/cisco/bigmuddy-network-telemetry-stacks

Juniper JUNOSgithub.com/Juniper/open-nti

Arista EOSvendor

Avendor

Bvendor

C

message broker (kafka)

TE alerts

applications

13

timeseries DB (influxDB)

dashboards(grafana)

OpenConfig models

Challenges in thinking “model-driven”

CLI habits are hard to break

implementations are often very different (occasionally for a good reason)

models for machines or humans

what to leave out of the model

how to test for compliance by an implementation

14

Summary

Models are critical for providing abstraction, durable APIs, and tractable checkability

topology, configuration, operational state, safety/security

OpenConfig operator group is developing and publishing data models for config and opstate -- built by users for users

as native implementations become available, OpenConfig will dramatically reduce proprietary integrations and simplify network telemetry, configuration, and qualifications processes

15