Computational Hardware

There are two ways to access BCG computational hosts.You can log into one of the open compute servers, nebula-1 through nebula-10. The nebula machines should only be used for interactive jobs. Note that you must be on the local network to log directly into any of these machines. See the status of logon servers here –

Alternatively, background or batch jobs can be run either on your own desktop machine or scheduled with condor.

Do not use the login hosts adhara or fornax for computation, large file downloads, or VS Code. They don’t have enough CPU or memory to be useful for that purpose, and it will only slow down other people’s sessions.

Very High Memory Servers

We have two high memory hosts, mastodon-1 and mastodon-2, with 500G and 40 cores a piece, and mastodon-3 with 80 cores. If you have jobs that have high memory requirements, please send a request to sends e-mail) with a very brief description of your need, and we will give you access to these machines.

GPU Compute Servers

We have two open-login GPU compute servers, and an additional two in HTCondor. Please send a note to to request access to these machines. We recommend using the open-login servers to test your code first, then use condor to schedule real GPU work.

Dedicated Condor Hosts

We have a machine pool of linux computers which do condor only, pulsar-1 through pulsar-40, and pulsar-70 through pulsar-94. No one has direct access to these machines.

Other Hosts

Your PI may have dedicated hosts available just for their lab.