Compute requirements

The compute requirements for analyzing MeerKAT data depend very strongly on the choice of software used, both for calibration and imaging. They also depend on the channelisation mode, whether the entire band is being used, or just a portion of the band is being used for spectral line imaging.

Assuming a full band 4K calibration, and widefield continuum imaging using CASA (without optimisation for parallel processing) will require a maximum instantaneous RAM usage of ~ 100GB, with an average usage of ~ 50GB. It will take ~ 24 hours to calibrate the 4K dataset, and a further ~ 24h to image using tclean (on a machine with 32 cores).

For a more modest desktop, we would advise maximising memory and storage. Another option is to use the SDP GPU-based imager.

Please have a look at our data volume calculator, which should give an idea of how much data needs to be handled. Depending on the specifics of the pipeline used, it is advised to have at least 3 times the MS size available in disk space i.e. if your spectral line measurement set is 10 TB, you will need at least 30 TB to calibrate and image the data.

South African-based PIs can apply to use IDIA (Inter-university Institute for Data Intensive Astronomy). To find out more, go to the IDIA site. Information with regards to datasets that are pushed to IDIA will have to be followed up with the IDIA Help Desk. For further IDIA access, related support and support for the ilifu Research Facility, please contact support@ilifu.ac.za.