In late 2014, VLCSI encountered restrictions in the way Karaage tracks usage information. Karaage had been developed based on requirements from VPAC, and the requirements of VLSCI were very different. In short, Karaage didn’t meet the needs of VLSCI.
VLSCI contracted Common Code to rectify the situation. This development effort was lead by Tyson Clugg, who developed new DB tables to make indexing and retrieval of the usage information more flexible and faster.
Unfortunately, after about 2 months of development effort, the funding ran dry.
The new DB models were developed, but the views were never updated to make use
of these new models. The work was copied to the karaage4 branch in the main
Karaage repository, but was not included in the recent Karaage 5 release.
This is a short description of some of the key changes that were made in the karaage4 branch, in case this project is ever resurrected.
A resource is a measurable attribute on a given machine. For example, it could be disk space or CPU time. A resource can have a scaling factor, so it can be scaled if required to ensure fairness between different machines with different capabilities.
A resource belongs to a resource pools. A resource pool defines a set of resources that can be grouped together. For example, disk space on /home for all clusters. The resource pool also defines the units for the resource (e.g. megabytes) and a function that should be used for aggregating results (e.g. megabytes).
All usage is initially written to the CPUJob table, same as currently. However it is also written to the new Usage table, that aggregates results and speeds up accesses.
This is a summary of the changes that were implemented, but never finished in Karaage 4. It is mostly from my memory combined with a quick review of the code, so errors might exist. I will update this post with new details as I remember them.
]]>