Faculty of Science Research Cloud - Cirrus
Cirrus is a OpenStack cluster, used to support Science Research.
Overview
What is Openstack?
Openstack is a cluster system, much like AWS EC2. There are Projects that have a quota of resources (CPU cores, RAM, Volume storage). From these resources, the Project users can create, run, and delete their virtual machines as they see fit.
How is Cirrus Different?
Cirrus supports Science research as cost effectively as possible. Here is what makes Cirrus special:
No over provisioning
The resources bought for a Project are available all the time, and are not shared with other Projects. This makes sure a rush to complete processing for a deadline is possible and not blocked by a queuing system.
Partial Cost recovery
Cirrus Projects are charged for the Resources they use. The Dean of Science covers the infrastructure and technical support costs.
Bulk storage system
Cirrus has a bulk storage system, which is not standard to OpenStack. This means there are two types of storage resources on Cirrus: Volume and Bulk. Bulk storage is much cheaper, and allows Science Projects to store and do research on petabytes of data.
On campus
Cirrus is in the University of Alberta campus data centers. Project data is quick to access, and does not leave Canada.
Moderate size
Cirrus is not meant as a HPC replacement. Canadian HPC resources, such as the Digital Research Alliance of Canada, should be considered as a Science researcher's first stop for large computing needs. Sometimes the national or local HPC services do not match with Science researcher's needs and that gap is what Cirrus may fulfill.
GPU Hosting
Cirrus can host GPUs. Cirrus Projects can purchase GPUs, and have them installed into Cirrus and attached to their VMs.
Cirrus GPU restrictions
- 2 PCIe slots wide
- 300W max power
- not open air fan cooled
History
Resources
Cirrus has the following resources (at time of publication):
| Resource | Amount |
|---|---|
| VM Servers | 52 |
| vCPU | 1752 |
| Memory (GB) | 13981 |
| Bulk Storage(PiB) | 4.1 |
Pricing
Cirrus pricing is based on 5 years of resource use. Shorter terms are available (1,2,3, or 4 years). The price for a shorter term is a simple fraction of the 5 year cost (eg. 1 year is 1/5th the cost). All charges are done through University speed codes.
| Units | Cost for 5 years |
|---|---|
| "CPU Unit" = 2 x vCPU + 16GB RAM | $450 |
| GPU Slot only (No CPU, No RAM) | $1000 |
| CEPH (100 GB) | $80 |
| Bulk storage (1TB) (scratch) | $76 |
| Bulk storage with snapshots (1TB) (work) | $152 |
| Bulk storage, with snapshots and replication (1TB) (home) | $304 |
Explination of price changes in 2026
Because of the shortage of computer chips (RAM, flash storage) the price of CPU units has gone up for the first time in Cirrus history.
Examples of Cirrus Projects
Coming soon.