Open mturilli opened 5 years ago
Just a quick comment on this one, w/o having precise measurements: we tested the REST interface in a different context (ATLAS/Panda), and found that utilization and TTX was comparable to RP being used stand-alone. In an integrated test with ATLAS, the performance was abysmal, but that was caused by an extremely slow task submission rate, which we don't have in the pull model used in the context of QCArchive.
This would be useful for RCT proposal too.