Benchmarking and scaling OpenStack in the enterprise

OpenStack has become increasingly important to enterprises of all sizes as its reach extends beyond Web 2.0/SaaS companies, such as Workday, Webex and PayPal. OpenStack is now a major element of both enterprise cloud computing and broader cloud initiatives, such as the new IBM Federal Cloud Innovation Center in Washington, DC.

In addition, OpenStack recently took further steps towards a broader appeal—over 1,500 steps in fact.

IBM, SoftLayer, and Mirantis have joined forces to further improve the performance of OpenStack. One of the most valuable outcomes of this collaboration is project Rally: a testing and benchmarking tool that measures OpenStack performance when running a variety of traditional enterprise workloads. To power these tests Mirantis has partnered with SoftLayer. Specifically, SoftLayer supplied essential infrastructure to conduct OpenStack benchmark testing using over 1,500 bare metal servers in order to improve scalability and perform other previously impossible OpenStack testing.

However, the ultimate goal of Rally is to go beyond testing OpenStack performance and provide the OpenStack community with tooling and reproducible scaling tests that will improve the upstream codebase and make upstream OpenStack both more relevant, and easier to adopt, in traditional enterprises.

Mirantis is developing Rally under OpenStack community guidelines, with code being available under the Apache 2.0 license. Rally natively integrates with other OpenStack community projects, such as Fuel for OpenStack deployment, Tempest for conducting tests, and Ceilometer for performing measurements.

OpenStack performance, tooling and reproducible scaling

Rally makes it possible to codify behavioral properties and associated workflows for a given application workload. It then simulates how such workloads run and scale on a particular configuration of an OpenStack cloud while measuring workload performance against an SLA.

“When using OpenStack for production workloads, enterprises can be confident that OpenStack performance will meet the SLAs specified for those workloads,” said Sesh Murthy, Vice President, Architecture and Technical Solution Design at IBM. “What is great about Rally is that it allows you to run a simulated workload on OpenStack for a period of time and report on workload performance against the SLA. This provides much-needed statistical data on infrastructure performance prior to movement into production.”

Mirantis recently announced its zero lock-in enterprise distribution of OpenStack hosted on SoftLayer and has been working with IBM to run the initial set of tests against OpenStack configurations and plug-ins supported by the Mirantis OpenStack distribution.

“To accelerate OpenStack adoption among mainstream enterprises, it is essential to provide customer confidence in the product by establishing credible performance benchmarks. Our initiative with IBM is an important step, and we encourage others in the community to join us,” said Adrian Ionel, Mirantis CEO.

How does Rally work?

Enterprises need reliable, cost-effective infrastructure capable of delivering to an SLA across workloads with a variety of behavioral and scaling properties. As an example, the most common use case for OpenStack in the enterprise has been around Dev/QA enablement. Assuming ten distributed teams working with a number of development and staging environments, each team will deploy an application to a given environment twice a day.

In OpenStack terms this means “X users provision Y number of VMs of a particular flavor Z times per hour concurrently.” This workflow can be codified with Rally, applied to a specific configuration of OpenStack and OpenStack performance can be measured against the SLA. Rally provides the workflow, user interface and workload simulation tests out of the box. An IT operations professional can follow a guided wizard to describe any workload in OpenStack terms and then run the simulation.

The video below demonstrates how to run simulated workloads on SoftLayer-hosted OpenStack clusters of 100, 500, and 1000 physical servers.

Mirantis, IBM and SoftLayer will reveal additional details of the Rally project in the coming weeks and at the IBM Pulse 2014 conference in Las Vegas.

Share
Comments: 1
Kirill Ishanov

About Kirill Ishanov

Kirill Ishanov runs Mirantis Labs and is responsible for research, evangelism and prototyping of OpenStack solutions and features aimed at making the platform easier to use and more reliable for enterprise adopters. Prior to heading up Mirantis Labs, Kirill ran OpenStack Services organization and was directly responsible for dozens of OpenStack implementations for some of the most prominent early adopters of the platform, including names like GAP, PayPal, Webex and others. Before jumping to OpenStack Kirill built IaaS and PaaS solutions for companies like eBay, Macy’s, Seagate and others. Kirill holds a master’s degree in Computer Science from Saratov State University. Outside of work, Kirill plays soccer, rides everything that has a shape of a board and plays video games with his daughters.
This entry was posted in All Posts, Government and the Cloud, open cloud, OpenStack and tagged , , , , , , , . Bookmark the permalink.

One Response to Benchmarking and scaling OpenStack in the enterprise

  1. Pingback: Benchmarking and scaling OpenStack in the enterprise | Dan Gorman's Technology News Aggregation

Comments are closed.