• Calendar.
• System Usage.
• Search.

• News.
• Help.
• Common Compute Currency.
• Hardware.
• Scientific Software.
• Compiling Codes.
• Running Batch Jobs.
• Performance Tools and Debugging.
• Data Analysis and Visualization.
• Data Management.
• Course Material.
• HPC Sources.

The new job accounting system is called the Common Compute Currency scheme.  The key feature of this new job accounting is that projects are given a global allocation which can be consumed on any system that a project's members have access to.  

It should be noted that nothing has changed in relation to the way projects are granted their allocation (this process still involves the rigorous technical, scientific and review-board procedure).  What has changed is that, unlike in the past, a project's allocation is now not bound to a particular computing resource but is global and can be consumed on any of the various computational systems at CSCS.

The overall design of the new scheme involves the following important changes:

1) Project allocations are now global across all CSCS computational systems;

2) On all systems (except Rosa) a local allocation threshold (quota) is enforced such that if a project's usage exceeds the local threshold, jobs will not run until the Project Office increases the local allocation threshold (quota).  This mechanism has been specifically designed to give the Project Office precise control over the way resources are consumed on the various systems;

3) Job priorities are influenced by the global usage not just the usage on a particular system (as before job size, requested wall-clock time, and time in the queue also contribute to a job's overall priority on a particular system);

4) The command-line tools ("monthly_usage » " and "sbucheck »") have been modified to report the global usage and usage breakdowns across the various systems;

5) The command-line tools no longer report usage in CPU-hrs but rather in Allocation Units (AU).  This is because usage is now no longer computed as the raw CPU-hrs consumed because this metric is becoming less and less relevant and also because the various systems have different architectures (e.g. some contain accelerators, some contain AMD CPUs while others contain Intel CPUs etc etc).  Furthermore some systems are less costly to run than others and over time these costs can change and hence the use of Allocation Units tries to reflect all these factors;

6) A new command-line tool "job_usage" has been created to report the exact breakdown of how a particular job (or list of jobs) was charged in Allocation Units.

****

Conversion Factors