|To:||Ben Gollmer <ben@xxxxxxxxxxxx>, XFS <linux-xfs@xxxxxxxxxxx>|
|Subject:||Re: Software RAID, a bit OT|
|From:||Seth Mos <knuffie@xxxxxxxxx>|
|Date:||Wed, 17 Jul 2002 21:17:21 +0200|
At 13:20 17-7-2002 -0500, Ben Gollmer wrote:
Hi storage gurus! This server is going to handle file sharing, e-mail, CVS, and a bug-tracking database for us. Our project has some rather large files so we need a good amount of storage space. I was planning to software RAID 5 the HDDs together for a total of 160 GBs. I have been enjoying XFS on my workstation but I know it has had problems with software RAID 5 in the past. Are these problems fixed now?
The only problems XFS had/has with raid5 are performance issues. They are not really problems in the sense that they break stuff. And performance fixes still occur every once in a while.
The real performance fix for raid5 is still in the works, untill that one is submitted everything still works and you can use it. I do it, I have a software raid5 with a internal log and I still don't consider it extremely slow in the sence that it is not usable. And the good thing is that it will only get faster in the future.
We also considered trying to grab another 80 GB drive from somewhere and do a RAID 0+1 (still giving us 160 GBs storage) but I don't know if Linux software RAID handles this well.
This works just fine. I already built one for testing and it performed and worked well. Just create 2 raid 1 devices and stripe those 2 raid 0 devices together.
So something like this:
md0 = hda + hdb md1 = hdc + hdd md2 = md0 + md1
Most of us run the -aa kernel tree on our workstations but I have no problem running SGI CVS kernels on the server if they are reasonably stable. Any input would be much appreciated :)
I am running the current CVS on some test machines but not on a production box yet. We decided to replace the broadcom gigabit cards(for intel e1000 cards) so we don't need the tg3 driver from 2.4.19 anymore.
Cheers -- Seth It might just be your lucky day, if you only knew.
|<Prev in Thread]||Current Thread||[Next in Thread>|
|Previous by Date:||TAKE - fixes a couple of btree assert trips in debug kernels, Steve Lord|
|Next by Date:||Re: TAKE - fix pagebuf locking bug, D. Stimits|
|Previous by Thread:||Software RAID, a bit OT, Ben Gollmer|
|Next by Thread:||Re: Software RAID, a bit OT, Danny Cox|
|Indexes:||[Date] [Thread] [Top] [All Lists]|