Supercomputing Division, Information Technology Center, The University Tokyo

Wisteria-A Fixed nodes

Overview

This course is able to meet needs that are difficult to handle when using the general group course. The following needs, for example, can be met using this course.

  • I wish to use the commercial programs that require license servers.
  • I wish to use dedicated HD to ensure data security.
  • I wish to lay down dedicated communication lines for data transmission.
  • I wish to have proprietary nodes to focus on calculations.
  • I wish to execute jobs that take a long time, and which cannot be handled using general batch job queues.

Application

Please download the application forms here.

Points to note

  • Only one set of Wisteria-A fixed nodes can be applied for.
  • The calculation resources that are contributed to the fixed nodes are limited to 6 set on the total.
  • A review will be carried out after application.
  • Please refer to the guidelines of the application form for other details.

Example of use: Application for 1 node

  • Number of tokens: 1 node(8GPU) x 360 days = 207,360 tokens
  • Use of batch job queues allocated for dedicated use
    • The total maximum number of nodes that can be used simultaneously within the group is 1.
    • Regardless of the number of GPU actually used, the token for time spent on 1 node will be consumed.
  • When 1 node are already in use for the dedicated batch job queues
    • When the same queues are assigned, you will have to wait until the required nodes are free.
    • If you wish to execute calculations at the same time, it is also possible to execute calculations using general batch job queues.
      • Tokens are consumed based on the number of GPU-hours used, and the number of tokens that can be used for the number of fixed nodes decreases accordingly.
  • If you wish to carry out calculations for more than 1 node
    • Execution using general batch job queues is possible.
      For example, if you wish to execute 4 nodes of calculations for four hours
      • Assign the general batch job queue (“regular”) and execute.
      • Number of tokens consumed: 32 GPU (4 nodes) x 4 hours x Coefficient of 3.0 = 384 tokens

Inquiries

Research Support Team, Information Strategy Group, Information Systems Department, The University of Tokyo
E-mail:E-mail address