Watcher, a Resource Optimizer for OpenStack: Plans for Newton

and Beyond

Austin Summit—April 28, 2016

ANTOINE CABOT (WATCHER PTL, b<>com)JOE CROPPER (IBM)SUSANNE BALLE (INTEL)

Watcher Mission Statement

Provide a flexible and scalable resource optimization service for multi-tenant OpenStack-based Clouds

Provide a framework to implement a wide range of optimization goals

Provide a pluggable architecture for new optimization algorithms and new data metrics

• Energy Aware optimizations e.g. outlet temperature

• Workload consolidations and re-balancing optimizations

“Out-of-box” value, albeit extensible!

https://wiki.openstack.org/wiki/watcher2

Many Contributors

3

Key Features for Watcher

Watcher provides:• Cloud optimization using VM live migration in case of imbalance detection • Granularity of optimization with multiple goals setting from a set of hosts to an entire

cloud• Evolutivity via its flexible plugin structure• « On-the-shelf » optimization strategies based on CPU, RAM and Energy

Watcher can run in:• “ADVISE MODE” for auditing before acting• “ACTIVE MODE” for always-on optimization (under development)• “VERBOSE MODE” for detailed optimization decision tracking (under development)

4

Watcher in the OpenStack Ecosystem

5

Watcher leverages services provided by other OpenStack projects

• VM live migration and resize• Metric collection• Power cycle bare metal nodes

Monitors the infrastructure and performs optimizations on-demand

Enables new ways for OpenStack administrators to reduce The Cloud’s TCO

Watcher

Nova(*)

Keystone(*)

Oslo(*)Ironic(*)

Ceilometer(*)

*Other names and brands may be claimed as the property of others.

Watcher Workflow and Maturity

Profile

Apply

Plan

Monitor

Optimize

Analyse

cost model constraints

Virtual machine IOPS,energy consumption,resources usage

Aggregate flows of events from the infrastructure and take action

Profile and predict virtual machine resource usage

Find trade-offs between objectives and constraints

Schedule actions such that all security, dependency and performance requirements are met.

Apply the optimal state where the infrastructure is utilized as efficientlyas specified in goals.

6 of 8

objective

3 of 81 of 8

Maturity level

6

Admin Control

Nova

Neutron

Cinder

Cloud Infrastructure

Ceilometer

Monasca

Horizon

Watcher Command Line

Watcher Dashboard

Wat

cher

API

Cluster Data Model

Cluster History

Time series databasesMetrics Collection

Open

Stac

k Bu

s

Watcher decision engine

Watcher actions planner

Watcher actions applier

Watcher DB

7

Workflow Description

1. Create a goal with a specific strategy• Goal is associated with a strategy in watcher.conf• Goal can be reused by another template

2. Create an audit template with a specific goal• Template can be reused by another audit

3. Create an audit with a specific audit template

• This audit will be triggered immediately in one shot or in a loop at a given interval

4. Audit will generate an action plan

5. Applier will take action manually or automatically 8

Overview of Watcher Decision Engine

9

Outlet Temperature Strategy Logic

Not found

Not found

Not found

Hypervisors with enough resource

Hypervisor with lowest outlet T Action Plan

VM Resource neededHypervisors

with outlet T

Hypervisors Over Threshold

Pick one VM of the Hypervisor

Filter

Sort

Telemetry from Ceilometer and data of hypervisors from Nova

Not found

Sort

10

Watcher’s Current State and Next Steps

11

Project Identified &Early Prototyping

Presented in Vancouver

Formed Initial Project Team

Mission Statement &

Key Uses Cases

Identified

Build Out Architecture Working Code

April

’16(M

itaka

)

Septe

mber

‘15May ‘1

5• Rebalance on server outlet

temperature and additional telemetry

• DevStack integration• And much more …

Novem

ber

‘14

• Sparked community interest

OpenStack compliant and ready for small

production deployment

Fall ’1

6(New

ton)

• Pluggable strategy• Additional strategies

coming from partners• Become part of “Big

Tent” for Newton

April

’17(O

cata)

OpenStack compliant and ready for large

production deployment

Mitaka Release: Framework Improvements

Setup all the OpenStack Tooling and processes for the project to run as an OpenStack project

Improved Watcher infrastructure• Watcher integration with ceilometer• Dynamic strategy load• DevStack plugin• Task Flow integration• Tempest integration• Multi-node support on the Gate• Threshold optimization• Watcher Dashboard

12

Mitaka Release: Additional Strategies

Outlet temperature strategy based on Thermal Telemetry

Rebalance VMs on physical servers strategy based on CPU/socket count/usage

Load consolidation strategy based on a heuristic algorithm which focuses on measured CPU utilization and tries to minimize hosts which have too much or too little load

Overload standard deviation strategy designed to integrate load balancing strategy based on a standard deviation algorithm

13

Plans for the Newton Release

Watcher scoring module• Standardize interactions between the Watcher Decision Engine and

Scoring engines, e.g., external machine learning tools or frameworks

Active mode for always-on optimization

Verbose mode for detailed optimization decision tracking

Provide efficacy indicator to the admin before running an optimization

Define a precise and extensible scope of resources to optimize (compute, storage, network…)

Provide more and more “value-added” optimization strategies14

Any Questions?

Want to learn more?!?

• Wiki : https://wiki.openstack.org/wiki/Watcher

• IRC : #openstack-watcher

If you are interested, we would love for you to get involved – come and see us!

Brown Bag session available on YouTube

15

Legal Notices and Disclaimers

Intel technologies’ features and benefits depend on system configuration and may require enabled hardware, software or service activation. Learn more at intel.com, or from the OEM or retailer.No computer system can be absolutely secure. Tests document performance of components on a particular test, in specific systems. Differences in hardware, software, or configuration will affect actual performance. Consult other sources of information to evaluate performance as you consider your purchase.  For more complete information about performance and benchmark results, visit http://www.intel.com/performance.    Intel, the Intel logo and others are trademarks of Intel Corporation in the U.S. and/or other countries. *Other names and brands may be claimed as the property of others. © 2016 Intel Corporation.