Mirantis, the pure-play OpenStack vendor, today announced dramatic new benchmark numbers for running an OpenStack cloud at scale to feed the growing appetite for demanding workloads on enterprise private cloud. The benchmark used the latest Mirantis OpenStack 4.0 (Havana) distribution running on CentOS 6.4, deployed in a high-availability configuration across distributed data centers - from start to finish in about 8 hours.
Using 350 physical servers on an IBM SoftLayer bare-metal multi-datacenter cloud, Mirantis OpenStack stood up 75,000 virtual servers, provisioned in parallel streams levels ranging from 100 at a time to 500 at a time. The benchmark was designed to show how quickly and reliably OpenStack could respond to on-demand, real world workload requirements for provisioning cloud resources, achieving a sustained a rate of 9,000 new virtual servers per hour for over 8 hours.
Key takeaways:
- Enterprises considering where to run large-scale workloads can rely on OpenStack for a private cloud with up to 75,000 virtual servers.
- Running Mirantis OpenStack on IBM SoftLayer provides highly responsive on-demand cloud resources, provisioned at over 9,000 virtual servers per hour.
- Mirantis OpenStack and IBM SoftLayer enable resilient multi data center cloud deployments.
"We built Mirantis OpenStack for scale and performance, and this benchmark with IBM shows that OpenStack is maturing quickly as a robust private cloud alternative to any public cloud offering today," said Adrian Ionel, CEO of Mirantis. "We're pushing OpenStack to its limits in harsh conditions that mimic real world production workloads and demanding customer use cases. Our goal is to give our customers reliable and transparent performance data for OpenStack at scale."
"OpenStack is experiencing explosive growth in the cloud market with more than 200 companies contributing code to the source," said Sesh Murthy, Vice President Advanced Cloud Services at IBM. "Customers embrace the proven performance and reliability of IBM Softlayer virtual and bare metal cloud servers and this new benchmark with Mirantis demonstrates just how powerful and flexible of a foundation we can provide with OpenStack for enterprise private cloud deployments at scale."
A detailed technical description of the benchmark configuration, tools and results is available at www.mirantis.com/openstackbenchmark/.
Technical Overview
Mirantis designed the test to demonstrate two key parameters of OpenStack: scale and speed of startup. Initiated from cloud controllers at a San Jose-based datacenter, Mirantis OpenStack was used to drive 250 parallel request streams for starting new virtual machines, across a network, to a complex of servers running in Houston, TX, more than 1,800 miles away. During the course of 8 hours, this distributed cloud infrastructure successfully built up 75,000 virtual servers running simultaneously on a remote infrastructure of 350 physical servers.
The virtual servers each included 1 vCPU, 64MB RAM, 2GB storage, with a CPU overcommit rate at 1:32 and a RAM overcommit at 1:1.5. Real-life workloads generally require much less simultaneous startup, so actual boot times per virtual server could be far faster. The test delivered an average sustained rate of 250 virtual servers bootstrapped every 98.8 seconds, growing almost linearly to the full population of 75,000 servers booted.