We've recently consolidated several STK 9714 DLT7000 tape libraries into one ADIC Scalar 1000 SDLT320 library with 12 drives.
We thought we'd identified all the potential bottlenecks that the backup system would face, and implemented the most robust solution we could come up with:
2 GB Etherchannel network connectivity
3 x 2 Gb FC connectivity between server and library
2 GB RAM
2 x 1.4 GHz Pentium III processors
72 GB Ultra 320 hard drives in a RAID 1
However, we've found that our nightly backup schedule has put a huge strain on our processor resources. We typically run ~400 jobs per night, and they begin within a 5 hour window, starting between 4:30pm and 9:30pm. So, at any given time, we could have ~100 jobs in an active state, consuming processor resources. (by "active", I mean: running, pre-processing, ready, or queued)
We know we could alleviate some of the processor load by spreading the jobs out over a longer window, but we also want to keep the backup window as short as possible for other administrative reasons.
We also don't want to set the Backup Exec services to run at a lower than normal priority.
Before we go out and spend thousands of dollars on a new 4-proc or 8-proc server, I'd like to know if anyone else is using a system that powerful to run their backups. Is that much power overkill, or do you think it would allow us to monitor our backups normally, and not have to wait several minutes before seeing a response to a click?
Thanks,
Jeremy
We thought we'd identified all the potential bottlenecks that the backup system would face, and implemented the most robust solution we could come up with:
2 GB Etherchannel network connectivity
3 x 2 Gb FC connectivity between server and library
2 GB RAM
2 x 1.4 GHz Pentium III processors
72 GB Ultra 320 hard drives in a RAID 1
However, we've found that our nightly backup schedule has put a huge strain on our processor resources. We typically run ~400 jobs per night, and they begin within a 5 hour window, starting between 4:30pm and 9:30pm. So, at any given time, we could have ~100 jobs in an active state, consuming processor resources. (by "active", I mean: running, pre-processing, ready, or queued)
We know we could alleviate some of the processor load by spreading the jobs out over a longer window, but we also want to keep the backup window as short as possible for other administrative reasons.
We also don't want to set the Backup Exec services to run at a lower than normal priority.
Before we go out and spend thousands of dollars on a new 4-proc or 8-proc server, I'd like to know if anyone else is using a system that powerful to run their backups. Is that much power overkill, or do you think it would allow us to monitor our backups normally, and not have to wait several minutes before seeing a response to a click?
Thanks,
Jeremy