News and Information

Notification of change of settings to limit the use of virtual memory of login nodes (gw, ngw)

The change of settings has been completed.

Dear HGC Supercomputer user,
June 18, 2013

Human Genome Center
Supercomputing Sevices

Thank you very much for your continued use of our center's services.

The virtual memory soft limits will be changed as shown below on gw, the login node of Shirokane1, and ngw, the login node of Shirokane2, to prevent them from becoming overloaded.

    Summary of Changes
      Before the Change
      gw (Shirokane1): 16GB
      ngw (Shirokane2): 16GB
      After the Change
      gw (Shirokane1): 2GB
      ngw (Shirokane2): 1GB
    Change Date
    2013/6/26

Following this change, programs that use more than 2GB of memory on gw, and 1GB on ngw, will no longer be able to run.

Please use qlogin if you would like to interactively run programs that use a larger amount of memory. Programs that use a large amount of CPU should also use qlogin.

The command below shows an example on how to use glogin. The command line arguments declare that it will use up to 8GB of memory.

$ qlogin -l s_vmem=8G,mem_req=8

For more details on how to use qlogin, please refer to pages 29-33 of the following training document: Using the Univa Grid Engine (in Japanese)

If you use qlogin on FreeNX, please refer to the following document for information on how to run X applications:

[Shirokane1] (in Japanese)

How to run X applications when you use qlogin (Shirokane1)

[Shirokane2] (in Japanese)

How to run X applications when you use qlogin (Shirokane2)

Login nodes are front ends for users of our supercomputer systems. High loads on login nodes can not only cause hindrance to the work of our users, but they can also stop users from being able to log in. Please ensure that heavy jobs are run on compute nodes, and not on login nodes.

Heavy jobs may be killed by the systems administrator without any prior notice. In the future, we are also planning to introduce processes that will automatically kill heavy jobs on login nodes.

We thank you all in advance for your kind cooperation in preventing our login nodes from becoming overloaded.

Index

Top of Page Top of Page

The University of Tokyo The Institute of Medical Science

Copyright©2005-2019 Human Genome Center