Lab Server FAQ

Who, where, what?

Our lab's server is called moneta and is a 20 cores/40 threads machine with 500GB of RAM.

Where to find/put data on moneta?

You have a home directory with very limited space. However, everything relevant to the group should be put on /proj/gibbons which is a large and shared lab resource that receives regular backups. Here an overview of some general resources there:

/proj/gibbons 
  > refs              # reference databases and artifacts
  > software          # software relevant to the lab (commercial solvers, etc.)
  > teaching          # training and reference materials
  > ...               # various projects and related data
  > 2025_my_project   # add your own (maybe start with the year for better sorting)

After a project is finished try to clean up large intermediate files (nextflow work directory for example) but always leave the raw data there. Please only edit files in directories you created or that were created/shared for you.

How do I connect to moneta?

You can only connect via ssh from the internal ISB network or when connected to the ISB VPN. The server is accessed via SSH which will be available when opening the Terminal on Mac, Windows or Linux. During the onboarding at the ISB you will receive your Linux username (the same as your short E-mail prefix) and password. With that you can connect to the server.

ssh user@moneta

Server is unresponsive but CPU usage is low

Within ISB most data is stored on dedicated file servers and accessed over the network. So usually all I/O will be limited by network speeds (the ISB network is pretty fast, so it's not that bad). However, especially when running running many threads or processes you can easily saturate network I/O (about 120MB/s max) which in the worst case will lock up moneta for everybody.

So here some tips to deal with that:

Use the local disk when possible moneta has about 512GB of local disk space available which will be faster than network I/O but is not part of any backup system and basically only exposes /tmp for write access. So a good strategy is to copy input data to tmp, perform your analysis writing all output to /tmp, and then move your final artifacts to the network locations such as /proj/gibbons.

Use the RAM disk if you need fast I/O and little memory If you want even faster file access you can also use the in-memory filesystem. This will store files in RAM which gives extremely fast I/O but will consume the used space in RAM. Thus, if you store 100GB of data on the RAM disk you will have 100GB less of RAM available for your workflow. To store data in the RAM disk simply write to /dev/shm. Please, remove all data from the RAM disk as quickly as possible after use.

If you use /tmp or /dev/shm please remove data from those locations in a timely manner. Long term storage should still happen in /proj/gibbons only as this has fail backs and a backup system set up. All of the data on /tmp and /dev/shm will be lost when the server reboots.

How do I change my password

You can change your password for moneta (and all other linux machines at ISB) by running the yppasswd command while logged in.