A colleague of mine and I finally convinced upper management that a similar methodology in our Citrix environment would be beneficial.
We've been using Blade servers in our Citrix farm for quite some time. With only two local drives, they were always configured as a mirrored set because, well thats just the way they've always done it. Noticing significant bottlenecks at the disks on some of the more heavily used application servers, we advocated breaking the mirror and making full use of each individual drive (page file location, temp directories, applications, etc.). We have enough servers in each application silo that we could lose one or two to a drive failure and continue operating without a hitch. We've just started testing but we're already seeing significant performance improvements. - Sean On Fri, Sep 11, 2009 at 9:07 AM, Jacob <[email protected]> wrote: > Actually, RAID0 does have its uses. > > We use RAID0 for the data on our web servers.. html files, images, videos. > > The o/s is on a single drive, the data is on a RAID0 with three drives. > Huge > performance gain. > > But, also... we have over 30 servers in the server farm. So if a hard drive > in the RAID0 crashes, the other 29 servers can take up the slack without > any > issues. I lose on average about two to three hard drives a year among the > 30 > servers. > > Now, if you have one database server and you use RAID0..................... > > -----Original Message----- > From: Ben Scott [mailto:[email protected]] > Sent: Friday, September 11, 2009 9:57 AM > To: NT System Admin Issues > Subject: Re: SATA RAID Performance > > On Fri, Sep 11, 2009 at 12:09 PM, Phillip Partipilo <[email protected]> wrote: > > Those are almost always software RAIDs, so your CPU would be doing all of > > the parity processing on RAID5 so performance there would quite suck. > > It depends on the workload (and just how crappy the fake RAID > implementation is). A lot of systems spend most of their CPU time > idle, so using that for RAID doesn't hurt you much. > > Over in the land of Linux, for example, OS software RAID performance > often blows even the best dedicated RAID hardware out of the water. > (It helps that Linux's RAID implementation doesn't suck anywhere near > as much as Windows' does.) I wouldn't do it on a web transaction > host, SQL box, etc., but if it's a file server, static web content > server, proxy cache, etc., sure. > > But given that the OP later states it's a gaming box, yah, I agree: > Steer clear from fake/host/software RAID. > > >There appears very low overhead striping in a RAID0 ... > > At the cost of doubling your chance of system failure due to disk trouble. > > These days, I wouldn't touch RAID 0 with a ten foot cattle prod. > > > when you created a RAID1 set, it would read the data in > > a striping pattern, essentially doubling read performance. > > Any decent RAID 1 implementation will do that. See above about > implementations that suck. ;-) > > -- Ben > > ~ Finally, powerful endpoint security that ISN'T a resource hog! ~ > ~ <http://www.sunbeltsoftware.com/Business/VIPRE-Enterprise/> ~ > > > ~ Finally, powerful endpoint security that ISN'T a resource hog! ~ > ~ <http://www.sunbeltsoftware.com/Business/VIPRE-Enterprise/> ~ > > ~ Finally, powerful endpoint security that ISN'T a resource hog! ~ ~ <http://www.sunbeltsoftware.com/Business/VIPRE-Enterprise/> ~
