After four years of zoom only meetings we finally had our yearly All-Hands Meeting in person in Madison Wisconsin.
At MIT our Tier2 computing center design is shifting from a fairly low power density hardware to high density CPU worker machines and dedicated storage servers. At the meeting in Madison we had an opportunity to learn what kind of worker nodes and storage severs other sites have or plan to deploy. As a result we do not envision any problems buying most recent servers with newest densest AMD EPYC servers (384/512 threads per 1 U for Zen4/Zen4c). Due to the large number of threads and the resulting massive amount of jobs (potentially even starting in parallel), those WNs would have high quality NVMe drives and best even 2 separate NVME drivers which reportedly brings the load to a level that the system can function.
Modern storage technologies such as CephFS and EOS were discussed in detail. We have considered CephFS as our most promising option for the future, but now CERN’s EOS also looks like an attractive option based on the talk given by Purdue team, that deployed EOS at their Tier-2 computing center. We will carefully evaluate pluses and minuses of each option and see what the best solution for our center at MIT is.
Another interesting topic that will affect all Tier-2 computing centers (at least in the US) was the presentation about the future for xrootd. The idea is that for some high priority large volume transfers dedicates routes will be used. While the idea is interesting it substantially complicates the logic in transfers and might lead to confusion and to adverse side effects at the sites.
Finally, we had a tour of University of Wisconsin Tier2 facility. That was the first time I had an opportunity to see another Tier2 installation.