Compute

Interactive High-Performance Computing Servers

About

Compute is a collection of large high-performance, shared memory compute servers for interactive use. These multi-socket machines host up to 88 Intel Xeon (Cascade Lake) CPU cores and 1.5TB of RAM. They provide access to various licensed software packages (MATLAB, SAS, STATA) and many open-source packages (R, RStudio, Sage Math, Apache Spark). All servers share the same user home directories which provides a total capacity of 0.5PB. EDR InfiniBand (100GB/s) is used to interconnect all machines.

These servers are the primary HPC resources for handling interactive calculations, applications that do not parallelize well and long-running calculations, that cannot be interrupted and restarted. So mostly projects with requirements, so that they are not permitted or not possible to run on the Owl's Nest HPC cluster.

This is the recommended resource to learn how to use Linux servers for computation, learn running, developing, testing, and benchmarking.

Compute has been upgraded!

To learn more about the hardware change, see the Hardware section.

Interactive Compute Server

compute is an interactive-use server which provides 88 CPU cores / 176 threads and 1536GB of RAM, making it ideal for short large-scale calculations. It is accessible from anywhere via SSH as compute.hpc.temple.edu.

 
 
Batch Compute Nodes

Eight batch operated compute nodes to supplement the compute server and with additional CPU cores and RAM. Their primary purpose is to run smaller jobs that do not require to be run interactively (e.g. Gaussian or parallel Matlab calculations), but are not parallel enough to justify using the Owl's Nest HPC cluster. Calculations are managed exclusively through the batch scheduler, but - unlike on Owl's Nest - node sharing is supported, i.e. one can request CPUs in single CPU increments. Running calculations across multiple nodes, however, is not supported. All of these systems are isolated from the outside network.

 
 
 
 
RStudio Compute Server

Finally, the rstudio machine enables users to launch interactive R-Studio sessions via their web browser by connecting to https://rstudio.hpc.temple.edu. This provides a convenient and platform independent access to the R statistical software. While R sessions run on this dedicated server, user data directories are the same as on all other Compute machines.

 

Usage Policies

The common policies for using IT resources at Temple also apply to the Compute servers. In addition, any use of the servers that is excessive (e.g. more than half of the available physical CPU cores and memory) or otherwise disrupting other users may result in temporary or permanent account suspension.