Run your Simulation on a high-performance Cluster!
Modeling and Simulation as a Service requires the presence of a flexible, robust and efficient computation platform. For that reason, the MARS Group operates an on-premise compute cluster with 140 CPU cores and 560 GB of main memory. The full cluster specs are as follows:
|CPU:||140 Cores (Intel Xeon 3.5 GHz & Core i7 3.0 Ghz|
|Memory:||Combined 560 GB ECC Memory|
|Storage:||30 TB usable network attached storage|
|Storage Network:||20 GBit SFP+|
|Compute Node Network:||2 GBit per Node|
|GPUs:||12x AMD FirePro D300|
To maintain flexibility and reliability while efficiently utilizing the provided computing resources, the MARS team decided to deploy the MARS Cloud as containerized applications and operate the cluster with the Kubernetes framework (https://kubernetes.io).
As the simulation complexity grows, more processing power is needed. MARS addresses this problem by abstracting from the physical hardware: Thanks to the growing support of Kubernetes by major cloud computing providers, a dynamic scaling to these providers or similar technologies is possible.