Cirrus: Building a Local Cloud for Researchers

Currently the gap in research computing between the desktop and large scale computing resources is typically filled by small clusters of servers run on an ad-hoc basis by individual research groups. Unfortunately they are also... [ view full abstract ]

Currently the gap in research computing between the desktop and large scale computing resources is typically filled by small clusters of servers run on an ad-hoc basis by individual research groups. Unfortunately they are also often poorly managed and put both data and security at risk.

The Cirrus research cloud in the Faculty of Science at the University of Alberta is designed to use cloud technology to consolidate these disparate clusters into a centrally managed facility. To ensure researcher buy-in we made purchasing Cirrus resources similar to purchasing standalone hardware, an approach well suited to highly variable research funding.

Researchers purchase life-long ownership of resource shares and over provisioning is tightly controlled to ensure that these resources are always available. Cheap, commodity hardware running OpenStack is also used to keep costs comparable to standalone server machines.

The advantages Cirrus provides to researchers are considerable: complete hardware maintenance and monitoring, a backup service, reconfigurable virtual hardware, rapid repurposing of resources and instant resource availability for small purchases. To ensure maximum flexibility researchers have full control and responsibility for their virtual machines.

Cirrus is managed by a combination of University IT support, who look after the hardware, and Compute Canada who install and manage OpenStack remotely. This ensures that Cirrus is compatible with large scale Compute Canada cloud resources so that Cirrus also acts as a stepping stone to use of large scale cloud facilities.