This technology deployment plan is a positive step forward. The approach for multi-purpose systems for a variety of science cases is impressive and will set the stage for future development in Canada.
- from the Compute Canada International Advisory Committee statement

GP1 — hosted by the University of Victoria

An OpenStack cloud system with approximately 6,000 CPU cores and 5 petabytes (PB) of disk for persistent storage. Both system and storage are being purchased, as of late June 2016, and will be deployed in summer 2016. An expansion to compute and disk resources of one third of the total expenditure is planned for mid-2017. Total CFI funding: $3M

Persistent storage, for all Stage 1 and 2 systems, is not inclusive of temporary/scratch disk, which will be purchased with the system. All core counts in this proposal are based on Haswell; actual core count will vary proportionally.

GP2 — to be hosted by Simon Fraser University

A large heterogeneous cluster with up to 25,000 CPU cores, nearly 1,000 GPU devices, and 15PB of disk for temporary and persistent storage. “GP” for GP2 and GP3 refers to General Purpose clusters, with various node types intended for diverse workloads. SFU is a destination for backups and deep storage. Installation and commissioning is expected Fall 2016. Total CFI funding: $8.35M

GP3 — to be hosted by the University of Waterloo

A large heterogeneous cluster with up to 19,000 CPU cores, hundreds of GPU devices, and 15PB of disk. Waterloo is a destination for backups and hierarchical storage management. Storage is being purchased, and the system is scheduled for purchase in October 2016. An expansion to compute and disk resources of one third of the total expenditure is planned for mid-2017. Total CFI funding: $7.8M

LP — to be hosted by the University of Toronto

A large parallel supercomputer with a balanced high performance interconnect, anticipating large homogeneous partitions with one or two node types, possibly including accelerator/manycore. Approximately 66,000 CPU cores will be purchased in mid/late-2017. 5PB of persistent storage is being purchased, and a further 10PB will be purchased with the system itself. Total CFI funding: $9.85M

Service Infrastructure Development

Research Data Management (RDM) and other software infrastructure investments in support of all systems and services, including the needs of CFI’s Challenge 1. Personnel and software expenses to be shared across the four Stage 1 sites. Total CFI funding: $1M

Infrastructure Systems

High-availability servers for critical infrastructure services have been deployed at SFU and Waterloo. All four sites are receiving new network routers to support a multi-site Science DMZ (see fasterdata.es.net), software-defined networking, and 100Gb connectivity. Funding included in site totals above.