Cedar

Cedar (GP2)

CEDAR is a heterogeneous cluster, suitable for a variety of workloads, located at Simon Fraser University. It is named for the Western Red Cedar, B.C.’s official tree and of great spiritual significance to the region's First Nations people. It was previously known as "GP2" and is still identified as such in the 2017 RAC documentation.

System evaluation is not yet completed as of November 2016. Anticipated specifications, based on SFU's RFP and bids received, include the following. This information is not guaranteed and might not be complete. It is provided for planning purposes.

Système de stockage

$HOME	Standard home directory Not allocated Small, standard quota Larger requests should be on $PROJECT
$SCRATCH Parallel High-performance filesystem	Approximately 4PB usable capacity for temporary (/scratch) storage. Aggregate performance of approximately 40GB/s. Available to all nodes. Not allocated Purged - inactive data will be purged
$PROJECT External persistent storage	Provided by the NDC. Available to compute nodes, but not designed for parallel I/O workloads.
High performance interconnect	Low-latency high-performance fabric connecting all nodes and temporary storage. The design of Cedar is to support multiple simultaneous parallel jobs of at least 1024 cores in a fully non-blocking manner. Jobs larger than 1024 cores would be less well-suited for the topology.

Types de noeuds et leurs caractéristiques

Noeuds de calcul de base:	Plus de 500 noeuds	128 Go de mémoire, 16 coeurs/socket, 2 sockets/noeud
Noeuds de calcul de type "large":	Plus de 100 noeuds	256 Go de mémoire, 16 coeurs/sockets, 2 sockets/noeud
Noeuds à grande mémoire 512	24 noeuds	512 Go de mémoire, 16 coeurs/socket, 2 sockets/noeud
Noeuds à grande mémoire 1500	24 noeuds	1.5 To de mémoire, 16 coeurs/socket, 2 sockets/noeud
Noeuds GPU de base	Plus de 100 noeuds	128 Go de mémoire, 12 coeurs/socket, 2 sockets/noeud, 4 GPUs/noeud, 2 GPUs/socket.
Noeuds GPU "large"	Environ 30 noeuds	256 Go de mémoire, 12 coeurs/socket, 2 sockets/noeud, 4 GPUs/noeud, Tous les GPUs sur le même socket
Noeuds à grande mémoire 3000	4 noeuds	3 To de mémoire, 8 coeurs/socket, 4 sockets/noeud

Tous les noeuds ci-dessus auront du stockage local.

Calcul Canada n'est présentement pas en mesure de spécifier le modèle de GPU. L'appel d'offre utilisait des NVidia K80 comme minimum.

The completed system is expected to have over 25,000 cores, 900 nodes, and 500 GPUs.

L'horaire de livraison et d'installation n'est pas encore déterminé. L'équipe d'acquisition est confiante que le système sera en production au début de l'année d'allocation, le 1er avril 2017.