Re: LVM Striping and RAID for performance?

2009-03-30 Thread Mark Post
>>> On 3/30/2009 at 3:23 AM, Rob van der Heij wrote: -snip- > But when you stripe the logical volume, you can't increase > the size anymore. That's not correct. A striped LV can be resized, there just has to be given enough new extents across the same number of PVs as there are stripes. So,

Re: LVM Striping and RAID for performance?

2009-03-30 Thread Rob van der Heij
On Mon, Mar 30, 2009 at 3:18 AM, Fred Schmidt wrote: > I have since found out that our SCSI disk will be allocated in 50 GB > LUN's and so we would have a stripe of exactly 1 disk for most of our > systems. So I guess that rules out LVM striping. For many installations it is a huge plus that you

Re: LVM Striping and RAID for performance?

2009-03-29 Thread Fred Schmidt
Thanks for the great responses. I have since found out that our SCSI disk will be allocated in 50 GB LUN's and so we would have a stripe of exactly 1 disk for most of our systems. So I guess that rules out LVM striping. Regards, Fred

Re: LVM Striping and RAID for performance?

2009-03-27 Thread Robert J Brenneman
Has anyone looked at a comparison of : LVM striping vs SW RAID0 vs dm-multipath and PAV I would suspect that LVM Striping and SW RAID0 will have similar performance characteristics since they are mostly using device mapper under the covers nowadays. LVM is more flexible, but is the code pat

Re: LVM Striping and RAID for performance?

2009-03-27 Thread Richard Troth
I see that you've gotten some excellent replies already. In our shop, we originally avoided striping on the Linux side (that is, within LVM) because we wanted to avoid a plaid or cross-hatch when the data hit the physical media. Some of us now believe that #1 that's not really a problem and #2 w

Re: LVM Striping and RAID for performance?

2009-03-27 Thread Tom Duerbusch
If you really want performance, the LVM should be stripped across RAID Arrays. And make sure you use multiple controllers. (i.e. in the DS6800 there are two controllers, each has half the cache. If you do LVM across both controllers, you have the potential of using all the cache, not just hal

Re: LVM Striping and RAID for performance?

2009-03-27 Thread David K. Kelly
cc Subject LVM Striping and RAID for performance? 03/27/2009 01:17 AM Please respond to Linux on 390 Port Would somebody please cl

Re: LVM Striping and RAID for performance?

2009-03-27 Thread Rob van der Heij
On Fri, Mar 27, 2009 at 6:29 AM, Mark Post wrote: > If you mean the disks in the storage array itself, then yes, because z/VM and > Linux can only start one I/O to a single device address at a time.  (Unless > you're using PAV.)  If your LVs are striped across multiple devices, then you > have

Re: LVM Striping and RAID for performance?

2009-03-26 Thread Mark Post
>>> On 3/27/2009 at 1:17 AM, Fred Schmidt wrote: > Would somebody please clarify - is LVM striping still of benefit to > performance if your disk is RAID? > > If so, why? If you mean the disks in the storage array itself, then yes, because z/VM and Linux can only start one I/O to a single d

LVM Striping and RAID for performance?

2009-03-26 Thread Fred Schmidt
Would somebody please clarify - is LVM striping still of benefit to performance if your disk is RAID? If so, why? Regards, Fred Schmidt NT Government, Australia -- For LINUX-390 subscribe / signoff / archive access instructi