Computing in HPC cluster on x86 CPU or GPU. Computing problem can be solved on one or several processors/nodes in parallel. HPC cluster is a remote self-service and available for several users simultaneously.
Cluster’s technical information is available here.
- one HPC cluster user account;
- 24/7 remote access;
- 1 TB space for short-term data storage;
- 5 GB RAM per CPU core;
- installed application software;
- briefing and technical support.
Procedure for using HPC cluster
Users are provided remote access to HPC cluster (maximum data transfer rate 1 Gb/s).
To receive access credentials, you must fill in the application form. After receiving the application, HPC Centre will contact you to inform on the following.
HPC cluster is available 24/7, and since simultaneous use of the cluster by several users is possible, the queue principle has been used, which means that a job is put in a queue in case the amount of resources needed for it exceeds the amount of available HPC cluster resources (processor cores or memory). The queue time depends on the number of waiting jobs and the amount of resources claimed by the user.
To balance the HPC cluster load, HPC Centre holds the right to change the amount of HPC cluster resources (like processor cores) available to a single user or the number of simultaneous jobs.
Maximum time spent on one job is 96 hours, after this period the job is automatically cancelled. If more time is needed for a job, the user can make a separate agreement with HPC Centre.
A user is assigned a folder /home/ for storing files and interim results. The maximum storage capacity is up to 1 TB, which may be increased pursuant to a separate agreement with HPC Centre.
If a job is executed on CPU, up to 5 GB RAM may be used per core. If more memory is to be used, more cores must be reserved accordingly.
Since the primary task of HPC Centre is to support and encourage academic research, it is forbidden to use the cluster for commercial purposes without a prior agreement with HPC Centre.
Users are responsible for using the software pursuant to the license. An external user (one who does not work or study at RTU) must provide the necessary software licenses oneself. In case of questions about licensing, contact HPC Centre.
Users must avoid activities that might interfere with the work of other users (queue bypassing, exceeding the reserved amount of resources, etc.).
Please, do inform HPC Centre if the cluster has been used for preparing any scientific publication or paper.
In case of scheduled cluster repairs or software update due to which the cluster is not accessible, HPC Centre will warn the users in email in time.
Service deadline for new HPC users depends on the amount and complexities of the preparation work (code adaptation, software installation, etc.). As a rule, it takes some 1–2 weeks.
Registered users have 24/7 access to the cluster and queue, while the availability of computing resources depends on the queue and the amount of resources claimed by the user.
Record of use and payment procedure
Recording of HPC cluster use is done by MOAB Accounting Manager. Note: only the job execution time is considered, not including the time spent in the queue.
Unit of CPU time is CPU core/hour, i.e., using one CPU core for an hour.
Unit of GPU time is GPU/hour, i.e., using GPU for an hour.
Note: To use GPU, at least one CPU must also be used which is recorded separately.
HPC Centre issues a monthly invoice for using the HPC cluster until the 10th date of the following month. The invoice is issued pursuant to the price list and sent to user’s email address together with detailed record of cluster use. The invoice must be paid within a deadline agreed upon with HPC Centre.