Re: [RE]Poorer performance with Raid0 then without?

On my fileserver I encountered a problem where adding a third hard drive
drove my hardware interrupts (under ANY os) through the roof.  The end
result was the performance equivelant of disabling DMA on all my drives.

I have never figured out why.

Just something to check for.

On 12/31/06, James Olson <[email protected]> wrote:
>
> What is the output of hdparm /dev/hde and hdparm /dev/hdg? Especially of
> interest is whether multcount, using_dma and readahead are set to on. They
> can be automatically turned off by the kernel if there are any seek errors,
> like can happen early in the FC5 boot sequence with raid0. I have a patch to
> redhat nash to fix that if that is the case.
>
>
>
>
>
>
> ---------[ Received Mail Content ]----------
>
> *Subject : *Poorer performance with Raid0 then without?
>
> *Date : *Sat, 30 Dec 2006 23:07:48 -0800
>
> *From : *"Listbox" <[email protected]>
>
> *To : *<[email protected]>
>
>
>
> Hi folks!
>
> I have been trying to use a
>
> RAID bus controller: Silicon Image, Inc. PCI0680 Ultra ATA-133 Host
>
> Controller (rev 02)
>
> to create a RAID0 workspace for MythTV, but it's not working right....
>
>
>
> I have two Seagate 160gb drives in a RAID 0 array on an SiI0680 PCI
>
> ATA raid controller. I formatted an XFS filesystem on
>
> /dev/mapper/sil_agbgdgbjfhei2,
>
> and set mythbackend to use it. ( My DVB card is a DviCO
>
> Fusion Gold 3,( Conexant CX23880).)
>
>
>
> When I watch live MythTV, I get terrible artifacts on digital TV, and slow
>
>
> frame rates on analog. When I reset mythbackend to use a ext3
>
> partition on a single drive ( /dev/hdc1 ), performance is acceptable.
>
>
>
> The RAID devices are hde and hdg, the/ partition is hdc.... to me this
> looks
>
> like all my physical drives give comparable throughput. This is what
> hdparm
>
> reports:
>
>
>
> /dev/hda:
>
> Timing cached reads: 1668 MB in 2.00 seconds = 834.10 MB/sec
>
> Timing buffered disk reads: 88 MB in 3.05 seconds = 28.87 MB/sec
>
>
>
> /dev/hde:
>
> Timing cached reads: 1680 MB in 2.00 seconds = 838.49 MB/sec
>
> Timing buffered disk reads: 168 MB in 3.03 seconds = 55.46 MB/sec
>
>
>
> /dev/hdf:
>
> Timing cached reads: 1688 MB in 2.00 seconds = 843.57 MB/sec
>
> Timing buffered disk reads: 164 MB in 3.00 seconds = 54.59 MB/sec
>
>
>
> /dev/hdg:
>
> Timing cached reads: 1696 MB in 2.00 seconds = 846.68 MB/sec
>
> Timing buffered disk reads: 162 MB in 3.03 seconds = 53.45 MB/sec
>
>
>
> I also tried upping the PCI latency of the RAID card with
>
> setpci -v -s 01:07.0 latency_timer=B0 # PCI0680 Ultra ATA-133 Host
>
> But this had no effect. The DVB card has a latency of 32(decimal).
>
>
>
> So, given the same hardware throughput, I would expect better performance
>
> with the XFS+RAID setup. This is not the case. I went to considerable
>
> trouble to get the RAID and XFS working on my Fedora 5 system, and it's
>
> pretty discouraging to see that it actually
>
> degrades instead of enhances performance.
>
>
>
> I just reformatted with ext3, and tried again, and got the same
> degradation.
>
> It's looking like the RAID is the problem, but I do not,NOT,NOT! want to
>
> un-stripe the disks.
>
>
>
> What else to try?
>
> Listbox
>
>
>
> _______________________________________________
>
> Ataraid-list mailing list
>
> [email protected]
>
> https://www.redhat.com/mailman/listinfo/ataraid-list
>
>
>
>
>  ------------------------------
>
> *Free Movies* <http://cinema.lycos.com/?if_Event=mail> 100's of Free
> Feature Length Films - Meet Friends, Watch Movies & Win!
>
> _______________________________________________
> Ataraid-list mailing list
> [email protected]
> https://www.redhat.com/mailman/listinfo/ataraid-list
>
>

_______________________________________________
Ataraid-list mailing list
[email protected]
https://www.redhat.com/mailman/listinfo/ataraid-list