Backup locally then network copy. Or ... network then Local?
sql-lover
Posts: 39
Hi,
I am having some SAN issues (writes) that it is indirectly affecting my backup jobs.
Current jobs are set to backup locally 1st (dedicated backup LUN for that) and then copy to network path. That's the order when the backup job was created. So I played around a bit and inverted the order. I put the UNC path on what suppose to be the local path (1st field). Then I specified the local path for what suppose to be the UNC for the network location.
The results confused me even more and not what I was expecting.
I checked with SQL monitor the network utilization and found that 2nd option (backup to network and copy locally) has a spike in network utilization, 40% for a 10GB database, but it keeps using SAN communication for until the job ends.
Now, when backing up to local path 1st then copy to network drive via UNC, the network utilization remains at 20% for the whole duration of the job. However, it is not until the end, than the iSCSI or SAN is used, I guess that for the verify step?
So ... what it's more convenient ? Also, does SQL Backup 7 do backups in parallel when using local then network option? Based on the network graph, it looks that's what the job is doing.
By the way, tweaking the SAN is not an option and it's something that was discussed already. However, average output stays consistent at 44MB/sec when using SAN. When using network, output goes up to 60 or even more.
I am having some SAN issues (writes) that it is indirectly affecting my backup jobs.
Current jobs are set to backup locally 1st (dedicated backup LUN for that) and then copy to network path. That's the order when the backup job was created. So I played around a bit and inverted the order. I put the UNC path on what suppose to be the local path (1st field). Then I specified the local path for what suppose to be the UNC for the network location.
The results confused me even more and not what I was expecting.
I checked with SQL monitor the network utilization and found that 2nd option (backup to network and copy locally) has a spike in network utilization, 40% for a 10GB database, but it keeps using SAN communication for until the job ends.
Now, when backing up to local path 1st then copy to network drive via UNC, the network utilization remains at 20% for the whole duration of the job. However, it is not until the end, than the iSCSI or SAN is used, I guess that for the verify step?
So ... what it's more convenient ? Also, does SQL Backup 7 do backups in parallel when using local then network option? Based on the network graph, it looks that's what the job is doing.
By the way, tweaking the SAN is not an option and it's something that was discussed already. However, average output stays consistent at 44MB/sec when using SAN. When using network, output goes up to 60 or even more.
Comments
Our recommendation is to backup to a local drive, and then "Copyto" to a network location. This is because the copyto option includes network resilience to retry the copy if there's a temporary loss of connection etc.
This sort of makes sense- you'd see an initial spike as the backup is done directly to the san (assuming you mean network communication to the san device?) and then the continual access following on until the end would be the file being copied.
This makes less sense to me- i'd assume the san usage to occur both with the initial backup and also the copy (assuming both the "local" and the "network" locations are actually on the san, just different sections of it)?
Redgate Software