Invest in Kamiak
The Kamiak cluster is managed under the general principles of condominium computing. Investors (individual researchers or groups of researchers) purchase compute nodes and storage from our service catalogue. These resources are then installed in Kamiak and administered by the Kamiak system administrators. Investors in the Kamiak cluster are guaranteed on-demand access to their purchase.
Additionally, unused compute resources are made available to the general user community through a backfill queue. However, this general access to resources can be preempted at any time by investors when they want to use the resources they have purchased. Investors are also given priority access to the backfill queue. The enhanced condominium model thus allows individual investigators or groups to access more computing resources than would be available by purchasing a stand-alone system.
Customers may purchase between 1 and 10 nodes without prior approval. Customers wanting to purchase more than 10 nodes must get the approval of the CIRC Advisory Committee.
Customers may install and use any software needed. If installation assistance is required, an average lead time of 2-4 weeks should be expected.
Nodes and Storage may be purchased by any member of the WSU research community regardless of their geographical location or unit affiliation.
If you would like to become an investor in Kamiak please visit our Service Desk.
Typically, new nodes are deployed within 6 weeks of clearing WSU purchasing.
Kamiak is designed to provide 24x7x365 availability with minimal unscheduled downtime. Kamiak may be taken offline for scheduled maintenance on a monthly or as-needed basis. Please see here for further information about maintenance schedules.
Standard Compute Nodes
Please note that these prices are estimates and are subject to change.
If you are planning on purchasing a node or applying for grant funding, please contact our Service Desk for up to date pricing and availability, or alternate configurations (e.g. different RAM configurations).
|Processors||Cores per Node||Memory||Local Disk||Networks||Warranty/Service Term||Budgetary Estimate|
|Dual Intel Xeon Gold 6338||64||512GB, 3200 MT/s DDR4||NVMe SSD||•10/25Gb Ethernet|
Special Compute Nodes
If you are interested in purchasing a non-standard compute resource, i.e. an accelerator equipped node or a large memory system, please submit a request to our Service Desk to get an up to date budgetary estimate. Below are some examples of large memory and GPU based specialty nodes.
|Node Type||Processors||Cores per Node||Memory||Accelerator option||Local Disk||Network cards||Warranty/Service Term||Budgetary Estimate|
|Large Memory - v1||Dual Xeon Gold 6338||64||1 TB (3200 MT/s DDR4)||NA||NVMe SSD||•10/25Gb Ethernet|
|Large Memory - v2||Dual Xeon Platinum 8368||76||2 TB (3200 MT/s DDR4)||NA||NVMe SSD||•10/25Gb Ethernet|
|GPU -v1||Dual Xeon Platinum 8368||76||1 TB (3200 MT/s DDR4)||2x NVIDIA Tesla A100 (6,912 CUDA cores & 432 Tensor cores each)||NVMe SSD||•10/25Gb Ethernet|
|5 years||Contact the CIRC team for details|
|GPU -v2||Dual Xeon Platinum 8368||76||2 TB (3200 MT/s DDR4)||4x NVIDIA Tesla A100 (6,912 CUDA cores & 432 Tensor cores each)||NVMe SSD||•10/25Gb Ethernet|
All users are provided with a home directory with a set limit. Users may also take advantage of the performant scratch file space to temporarily store data. If you would like additional long term “Project Storage”, this can be purchased through the Service Desk in 500GB increments.
Please note that storage rentals are in gigabytes (GB) which is 10003 bytes, not gibibytes (GiB) which is 10243 bytes and thus slightly larger. To see your storage usage be sure to use “df -H” which shows gigabytes and terabytes.
|Storage Type||Usable Storage Amount||Funding Source||Yearly Cost|
|Project Storage||500GB unit||Federal grants/contracts||$45 /year|
|Project Storage||500GB unit||All other sources||$49 /year|