> Hi Michael, thanks for answering.
>> On a full, it looks like my systems generally get around 6Mb/s, and
>> to 1Mb/s on an incremental -- comparing the files may not take a lot of
>> bandwidth, but it does take resources on both sides, so what you're
>> may be normal for an incremental, I don't know.
>> My backup server is a Q6600 (4x2.4Ghz) system running Linux kernel
>> and the file system is xfs on RAID6.
>> On the client side, I selected a client to look at at random, it happens
>> to be running FreeBSD 9.1 and the filesystem is ufs.
> Based on the above, I seem to be on the very slow side, incrementals
> The main difference I see (apart from the RAM that was pointed out by
> Jeff), is the file system. Whilst I am using ext3, you are on xfs. Some
> report xfs performance increase is noticeable. Is that your experience
> and would you think that I should move mine to xfs as well?
> I have started another backup last night to gather yet another set of
> numbers to compare with the results you've kindly provided.
> Many thanks
When I migrated to RAID6 a couple of years ago, I tested xfs, jfs, ext3/4,
and in my application, xfs was the performance winner, followed closely by
jfs. ext3/4 had a relatively poor showing -- your mileage may vary, of
course; the majority of our backups are Windows systems using an rsync
client. (I'm afraid I'm traveling, so I haven't got more concrete numbers
handy, though I did report them on this list at the time.)
My tests were done on a 6TB RAID6 array, with 2G RAM (the system has
subsequently been upgraded to 8G; it runs a 32-bit kernel with PAE.)