When real garlic gets old, it gets moldy and stinky. Fortunately our garlic (an AFS server for OSG file space) didn't get stinky, but it did get too slow for the job. So we've upgraded from a lowly 2 GHz Pentium4 system to a spiffy dual 3.0 GHz Xeon system with 4 GB of memory.
Rosemary is now the "project/data AFS file server" for UW-HEP. (It's the spiffy new server with gigabit ethernet and 4.5 TB of very fast Apple Xserve RAID disk.)
Garlic has been relegated to the role of the "junk file server." As such, it's housing the osg and vms volumes.
Lastly, anise is housing home directories for accounts starting with A-M, and thyme is housing home directories for accounts N-Z.
Our AFS "junk server" garlic has been overwhelmed with OSG jobs lately so I cranked out "rosemary.hep.wisc.edu"--the AFS file server for the CMS Tier2 Data Center project. It's a 1U dual 3.0 GHz Xeon, 4 GB mem with 4584 GB of AFS RAID file space via an Apple Xserve RAID. So we're migrating from a single CPU with 10/100 ethernet to two 33% faster CPUs with gigabit ethernet and from ~ 30 MBps disk read throughput to ~ 136 MBps disk read throughput. I think the thing will be a bit more speedy--even when not going down hill with a tail wind.
Our AFS "junk server" garlic has become unstable: twice this afternoon it stopped working--the RAID controller suddenly reported two drives missing.
As a result, we lost the root.atlas AFS volume. I started restoring it a few minutes ago.
Also, I've started moving the other volumes to other servers so we can take garlic out of service.