I'm doing some lab tests of VADP backups, specifically relating to dense
filesystems. I noticed some odd results, and thought I'd see if anyone else
has noticed the same.



My test system has 1.8 million files, average file size of 10K.

I'm seeing a much slower backup when FLR is enabled than when it is not;
that is true when counting the job as a whole, or when just counting tape
time.



As I'm running it, my VADP Proxy has save.exe running at 13% CPU, with a
backup speed of 1-2MB/second; while its possible that there is an issue in
my lab VM environment, I would think that the actual data transfer phase
would be the same whether or not it was using FLR.

(And just to forestall any questions about the configuration: save.exe is
spurned off by nsrvadp_save.exe to do the actual data transfer. I see
nsrvadp_save and nsrvddk running on the VADP proxy for around an hour
before the actual data transfer starts)



1.8million files, 10K average size



Type             Level   Start time    Tape start time   End Time

VADP (no FLR) Full 5:00                                                 6:04

VADP (FLR)      Full  7:00               8:08                         11:07



Environment:

NSR 7.6.3.7, Windows x64.

VMware 4.0, VADP proxy is virtual using hotadd

Backup command: nsrvadp_save.exe

Application Info: VADP_HYPERVISOR=vcenter;VADP_VM_NAME=host1;
VADP_DISABLE_FLR=yes/no (depending on which test)

Proxy Backup: VMware, proxy host: VADPProxy:vcenter (VADP)



Has anyone else noticed this, and did you investigate the reason/resolution?



Thanks,



Dave







Dave Gold

Sr. Technical Consultant

Cambridge Computer Services

dgold AT cambridgecomputer DOT com

781-250-3260

*Celebrating our 20th Anniversary*