CBT and multiple replications/backups

[pufferdude]

I've read the FAQ that says that CBT will handle backup and replication of the same VM without getting confused as to what blocks have changed from each backup/replication's perspective, but are MULTIPLE replications supported? Here's my situation, and I'm seeing unexplained results:

1. 80GB VM on SAN backed up daily to local NAS.
2. Same VM replicated every 2 hours to local ESX storage. This is my "oh crap, the SAN died" recourse for this critical VM, so at most we lost two hours.
3. Recently I added an ADDITIONAL replication of this VM daily to a remote ESX box, across our 20Mbps WAN link. This is my off-site "oh crap, we lost the datacenter" backup of this VM.

Here's the unexplained part... The initial replication (#3 above) of this VM across the WAN took about 7.5 hours. 24 hours later it took only 6 minutes, and the next night a similar short time. However, the LOCAL "every 2 hours" replication (#2 above) is taking about 3.5 minutes to complete EVERY TWO HOURS, to local (fast) NAS storage. I find it extremely hard to believe that the daily WAN-replication job (#3) is only taking 6 min to complete with 24 hours of changes while the every-2-hr local-replication (#2) job is taking 3.5 min to local storage... which makes me wonder if CBT really isn't working as advertised and the *WAN* replication job is actually seeing only the changes since the last LOCAL replication? Is that possible? Am I making sense? The math just doesn't make sense, and makes me wonder if multiple replications (plus a daily backup) of the same VM really screws up and CBT really can't track changes for each job separately.

Can anyone confirm or deny this? Any way to know for sure how CBT is handling changes for these three separate jobs on the same VM?

[Gostev]

Yes, multiple replication jobs are supported too. Job type does not affect anything at all (just different target). Also, I do not see anything unusual in the above numbers.

[tsightler]

A couple of points. Replicating every two hours will always generate more data than replicating every 24 hours. Things like the swap file, or log files might change the same block multiple times every hour, which means that data has to be backed up on each cycle, however, it only has to be backed up once for the daily backup. There's also a lot of overhead in every replication cycle, time spent connecting to vCenter, sending the agents, taking snapshots, removing snapshots, etc, etc. It sounds like your system has only limited changes so very little of the time is actually being spent transferring data.

I'd say what you're seeing is pretty much just "normal", but my best suggestion would be, if you're worried about it, just test it. There's certainly no issue with having multiple replication jobs with CBT, we do it all the time.

[pufferdude]

Thanks for the sanity check... I guess I wasn't accounting for all of the overhead! I sure wish Veeam reported ACTUAL bytes transferred during a backup/replication instead of just the fun-but-not-very-useful effective throughput numbers.

[Gostev]

Me too this is like my own number 1 enhancement request!