Cerebro is a cluster of twenty dual-processor six core Intel Xeon X5650 based servers and eight dual-processor eight core Intel Xeon E5-2650 servers. Each X5650 machine has twelve cores and 24GB of RAM and each E5-2650 has sixteen cores and 64GB of RAM. They all run Linux.
Cerebro can only be accessed by sshing into the head node, whose external name is cerebro.ch.private.cam.ac.uk. All work is done from there; there is no need to log into compute nodes. Cerebro uses the local Admitto service, so you log in with the same password as on the workstations.
Homespace is on a disk array attached to the head node. The /home filesystem is 1TB in size and has user quotas which are set to be very restrictive (25GB). It is backed up and various levels of incremental are kept. /home is shared to all nodes on the cluster's internal network, so your job sees the same home directory wherever it is on the machine. It's important to remember that from a compute job's point of view accessing this directory is extremely slow, especially if all the nodes are trying at once. Compute jobs should always write data to a local disk if possible, and copy it back to /home at the end.
There is a much larger space available on /filestore ; every user should have a directory there with a quota of 500GB. It is backed up and various levels of incrementals are kept. This space is located on a filestore node and is still slow to access compared to local disks on compute nodes.
There is also a shared scratch filesystem /sharedscratch in which you will have a directory. These are not backed up. They have a quota restriction but it's generous and it is expected that most people will stay well within it. They have the same speed issue as /home and /filestore.
Each node also has a local /scratch filesystem on which the queueing system will create you a directory when you use the node. These filesystems are about 900Gb in size with no quota restriction and are the most appropriate place for your jobs to write temporary files during a run. They are local to each node and so considerably faster than the NFS-mounted /home and /sharedscratch. Please clean up files on /scratch when you are done with them; see the queueing documentation for how to find out which node's /scratch to look at. All of the node /scratch directories are accessible under /nodescratch on the head node. The system uses an automounter so the directories only appear when you try to access them. For example to see the /scratch from node node-0-13 you need to type something like 'ls /nodescratch/node-0-13' .
A variety of compilers and libraries are installed. Like most local Linux machines cerebro has the modules environment to allow you to switch between different compilers and libraries. Use the module avail command to see what the other options are, and edit the version you want into your ~/.bashrc file.
All compute jobs should be run through the queueing system. The queueing system will run each job on a set of free compute nodes, copying the output back to a user-specified file at the end of the job. The queueing system is SLURM; this will be familiar to users of the other theory sector clusters, but please note that the available partitions are not the same on every machine. Read the generic instructions for instructions on how to use SLURM and Cerebro's queueing setup for details of the queue setup if you already know SLURM. To run parallel jobs you will need to use the appropriate job launcher; see Cerebro's queue and parallel documentation for details.
