Would adding a dedicated ZIL/SLOG (what is the difference between those 2 exactly? Is there one?) help meet your requirement?
The idea would be to use some sort of relatively large SSD drive of some variety to absorb the initial write-hit. After hours when things quieit down (or perhaps during "slow periods" in the day) data is transparently destaged into the main disk-pool, providing you a transparent/rudimentary form of HSM. Have a look at Adam Leventhal's blog and ACM article for some interesting perspectives on this stuff... (Specifically the potential "return of the 3600 rpm drive" ;-) Thanks -- mikee ----- Original Message ----- From: [EMAIL PROTECTED] <[EMAIL PROTECTED]> To: Tharindu Rukshan Bamunuarachchi <[EMAIL PROTECTED]> Cc: zfs-discuss@opensolaris.org <zfs-discuss@opensolaris.org> Sent: Wed Jul 23 11:22:51 2008 Subject: Re: [zfs-discuss] [zfs-code] Peak every 4-5 second On Wed, 23 Jul 2008, Tharindu Rukshan Bamunuarachchi wrote: > 10,000 x 700 = 7MB per second ...... > > We have this rate for whole day .... > > 10,000 orders per second is minimum requirments of modern day stock exchanges > ... > > Cache still help us for ~1 hours, but after that who will help us ... > > We are using 2540 for current testing ... > I have tried same with 6140, but no significant improvement ... only one or > two hours ... Does your application request synchronous file writes or use fsync()? While normally fsync() slows performance I think that it will also serve to even the write response since ZFS will not be buffering lots of unwritten data. However, there may be buffered writes from other applications which gets written periodically and which may delay the writes from your critical application. In this case reducing the ARC size may help so that the ZFS sync takes less time. You could also run a script which executes 'sync' every second or two in order to convince ZFS to cache less unwritten data. This will cause a bit of a performance hit for the whole system though. You 7MB per second is a very tiny write load so it is worthwhile investigating to see if there are other factors which are causing your storage system to not perform correctly. The 2540 is capable of supporting writes at hundreds of MB per second. As an example of "another factor", let's say that you used the 2540 to create 6 small LUNs and then put them into a ZFS zraid. However, in this case the 2540 allocated all of the LUNs from the same disk (which it is happy to do by default) so now that disk is being severely thrashed since it is one disk rather than six. Bob ====================================== Bob Friesenhahn [EMAIL PROTECTED], http://www.simplesystems.org/users/bfriesen/ GraphicsMagick Maintainer, http://www.GraphicsMagick.org/ _______________________________________________ zfs-discuss mailing list zfs-discuss@opensolaris.org http://mail.opensolaris.org/mailman/listinfo/zfs-discuss _______________________________________________ zfs-discuss mailing list zfs-discuss@opensolaris.org http://mail.opensolaris.org/mailman/listinfo/zfs-discuss
