New Anansi cluster for notebooks and test/debug jobs#
We are happy to announce the immediate availability of a new cluster designed for interactive use (i.e. notebooks) and short jobs for testing and debugging purposes. The name of this new cluster is Anansi.
Anansi lives next door to its bigger sister Hydra and both share the same storage system, so your files and data are readily available on both clusters. Their similarities end here though, as Anansi is specifically designed for short jobs and quick access.
The scale of Anansi is much much smaller than Hydra. It is starting with:
32 Intel Broadwell CPU cores
4 NVidia GeForce 1080 Ti
512 GB of system memory (RAM)
The hardware is slightly older than what can be found in Hydra, but this configuration of resources already allows to test many real use cases like serial jobs, non-MPI parallel jobs on CPU and jobs on GPUs. You will find the exact same software modules available on both Hydra and Anansi.
Moreover, Anansi uses a totally different allocation policy and its resources can be shared among running jobs. This means that jobs will not queue whenever its 32 cores are in use, or all 4 GPUs are in use. More jobs will be able to start and use those same resources. The reason for this approach is that interactive and debugging workloads are not intense in resources and have frequent idle periods. Therefore, multiple jobs/users can use the same CPU or GPU in such scenario. In total the capacity of Anansi is 128 single-CPU jobs and 16 single-GPU jobs. Finally, to further avoid queue times, the maximum walltime of jobs in Anansi is limited to 12 hours.
Please check our documentation for more information on:
If you have any comments or questions about this update, please contact us at VUB-HPC Support.