Understanding memory settings

Most work done by the CLC Science Server is done via its java process. However, particular tools involving de novo assembly or mapping phases (e.g. read mappings, RNA-seq analyses, smallRNA analyses, etc.) use external native binaries for the computational phases.

Java process

For the grid worker java process, if there is a memory limit set in your clcgridworker.vmoptions file, this is the value that will be used. See in Other options.

If there is no memory setting in your grid worker's clcgridworker.vmoptions file, then the following sources are referred to, in the order stated. As soon as a valid setting is found, that is the one that will be used:

  1. Any virtual memory settings given in the grid preset, or if that is not set, then
  2. Any physical memory settings given in the grid preset, or if that is not set, then
  3. Half of the total memory present, with 50GB being the maximum set in this circumstance.

Please note that any nodes running a 32-bit operating system will have a maximum memory allocation for the java process of 1.2 GB.

External native binaries

For the computationally intensive tools that include a phase using a external native binary, (e.g. de novo assembly, and jobs involving mapping phases (e.g. read mappings, RNA-seq analyses, smallRNA analyses, etc.)), the external native binary phase is not restricted by the amount of memory set for the java process. For this reason, we highly recommend caution if you plan to submit more jobs of these types to nodes that are being used simultaneously for other work.