In the past year, the OSG delivered nearly 90 million CPU hours to so-called opportunistic virtual organizations (VOs). These opportunistic VOs are ones which do not have large computing resources of their own and thus rely on opportunistically running on resources not owned by the VO. The largest of these, the OSG VO, which mainly consists of individual researchers from a wide range of disciplines, obtained over 60 million CPU hours since July 2013.
Expanding the number of available opportunistic CPU hours on the OSG is a priority for the OSG Executive Team. As the production analyst for the OSG, I, along with OSG project manager Chander Sehgal, have been working on characterizing the opportunistic ecosphere and identifying avenues to expand it. This task is complicated by the fact that each site on the OSG has its own policies and procedures regarding opportunistic use, resulting in not one, but over a hundred different “opportunistic ecospheres” to study.
After studying opportunistic availability and use at the OSG over a period of several months, we are now in contact with managers at several large OSG sites to understand how we can more efficiently harvest opportunistic cycles at these sites. We are also working on devising recommendations for OSG VO users on how to tailor their jobs to better take advantage of opportunistic cycles that may be available.
By working on these two projects and taking into account natural overall growth of the OSG, our aim is for the OSG VO alone to harvest 100 million CPU hours from the OSG in the coming 12 months.
~ Bo Jayatilaka