I am running DP 6.2 on an HP-UX cell manager (Core patch PHSS_42856) that is the media server (patch PHSS_42860) for our HP-UX, Windows, and ESXi environment. I have nine ESXi 5.0 hosts that are all very similar in age, CPU technology, and memory. They are located in two IBM blade centers. I have configured host backups to a Windows backup host (Core patch DPWIN_00567, VE integration patch DPWIN_00566) which passes the data to the cell manager/media server. The end device is the same SDLT drive in an MSL5052 library via FC connection.
The backups of ESXi blades in one blade center run on average 100GB/hr. While the blades in the other blade center run on average 20GB/hr. (Throughput calculated by total data backed up divided by total duration of backup) Networking configurations on EXSi and Cisco infrastructure is all identical.
My network admin and I have already discussed LAN limitations but that doesn't appear to be an issue (one of the slow hosts asked for more bandwidth and received it). None of the ESXi hosts are CPU or memory limited.
I would appreciate any ideas on what I can check to identify bottlenecks.
Do you perform NBD (network) or SAN backups of the ESXi datastores? Another possible root cause could be the originating disk subsystem used by ESXi or just the Data Protector configuration itself. I usually see even more throughput for a single VM backed up using NBD transport in a VEPA session. (to LTO3-5 or disk) How many VMs are backed up at once? What about concurrency on the drive and how many drives are used at once? If the ESXi storage is fast, your SDLT drive could be the bottleneck. If it is as old as your MSL5000 then I would recommend to backup to a null device on the media agent. If the throughput is much higher, your drive is just to slow.
--- Please assign a KUDO to this post, if you find it useful.