Hi All,

Any other CIFS users seeing very poor CIFS performance when accessing large numbers of small files?

The specific problem is with across network backups using ARCServe. We were seeing 70MB/min to a DLT3 drive when running OnTap 5.1D3, but upgrding to 5.2.1 has dropped that back to 20MB/min.

When I backup large files (256MB core files for example) I'm seeing backup throughput of ~130MB/min which is pushing the DLT drive close to its limit (150MB/min with 2:1 compression).

In general the performance for large files is better in 5.2.1, but Netapp seem to have broken the small file throughput.

I've been through the usual tests: check duplex, swap network cables, even moved both backup host and the filer to another switch. Given the throughput on large files, I can't see it being a network problem. Neither the switch nor the filer interface are showing large number of errors.

Backup performance from other NT servers hasn't dropped off.

We backup from a snapshot. The small files (source files mainly, we're a SW house) are in quota trees using mixed security. There's quite a complex ACL system in place and some files may have 6-8 NT groups as part of the ACL.

cifs.oplocks.enable is on and cifs.lvl2_oplocks_cap is off, which hasn't changed since the upgrade.

I have a support call open with Netapp (Log ID: 52724.00000) and have provided them with a 60s network trace of the traffic between the filer and backup host when the backup throughput has gone way down and there's not much other filer usage, but they have yet to get back to me.

Can anyone think of anything I should check for?

Thanks, GB