Re: [rdiff-backup-users] Some idea of timings...

[dean gaudet]

On Sun, 16 Mar 2003, Spicer, Kevin wrote:

> HARDWARE 1.5 GHz intel Pentium IV (maybe dual processor...), 1GB memory,
> Compaq Smart Array (RAID - probably RAID 5)  OS Red Hat 7.2 (I have good
> reasons for not using 8 or later yet - a third party application that is
> only supported under RH7.2 by the vendor) 100Mbs HD link between servers
> (which will be on different sites)
>
> Area being backed up will be 100GB on the RAID array (at both ends)
> about 1 Million files.  Probably a maximum of 5% of these files will
> change in any one day.

my primary server is dual piii 750 w/1GB RAM... 4x maxtor 7200rpm using
linux sw raid5.

the backup server is dual piii 500 w/512MB RAM... 7x maxtor 5400rpm using
linux sw raid5.

backup occurs across a pacbell DSL line (1.5 down, but the 128k up is
painful for rdiff-backup because that's the direction the signatures are
communicated in.)

i'm still using 0.11.0.

--------------[ Average of 87 stat files ]--------------
ElapsedTime 16099.14 (4 hours 28 minutes 19.14 seconds)
SourceFiles 744501.229885
SourceFileSize 79312563447.1 (73.9 GB)
MirrorFiles 743876.91954
MirrorFileSize 79182911479.2 (73.7 GB)
NewFiles 3262.17241379
NewFileSize 478664919.276 (456 MB)
DeletedFiles 2637.86206897
DeletedFileSize 358683806.299 (342 MB)
ChangedFiles 2905.3908046
ChangedSourceSize 3864098448.68 (3.60 GB)
ChangedMirrorSize 3854427593.76 (3.59 GB)
IncrementFiles 8806.42528736
IncrementFileSize 429379403.0 (409 MB)
TotalDestinationSizeChange 559031370.897 (533 MB)
Errors 307
--------------------------------------------------------

errors are due to a frequently changing part of the disk (i don't do
snapshots on this system yet.)

if i had to estimate the bottlenecks in this system i'd say it's the
primary server -- its cpu is usually maxed out, and there's substantial
other i/o going on for mail/web/services.  and at times the 128k uplink is
probably the bottleneck.  the backup server isn't typically doing anything
else.

-dean