Asked to help collecting information to add to the ATLAS presentation at the next LHCOPN/LHCONE meeting.
A Couple of plots available showing the Manchester upgrade (local prometheus vs ddm dashboard. But there are few more info and plots we can collect.
- For representative sites (e.g. 1k slots, 5k, 10k, 100k) network plots ,
FTS and internal (split).
→ important to know the external network config.
- Sites with Ceph can the internal storage replication traffic be monitored?
Looking at RAL, Lancs and Glasgow here
-
CVMFS cache traffic manchester squids don't go beyond 30MB/s with an average of 7-8 MB/s so ~ 60-70 Mb/s is negligible.
I assume the same for everyone?
-
StorageLess sites Oxford and Bham plots.
The picture James presented at gridpp48 was a bit foggy. It is not clear if the caches work or they are a nuisance.
But to me the main problem is to understand if we have a monitoring we can reliably look at when there are changes such as enabling disabling VP.
Or adding removing the cache between the remote node and the WNs.