MD1000 and PERC5/E I/O bottleneck

[Eric Yablonowitz]

Nope.  Nowhere close.  PCI-e x8 has a bandwidth of 32Gbps.

http://www.dell.com/content/topics/global.aspx/vectors/en/ 
2004_pciexpress?c=us&l=en&s=corp

On Jun 4, 2007, at 10:00 AM, Vincent Janelle wrote:

> You're hitting the upper limits of pci-e right now, its limited to 2.5
> gigabits per second.. I think.
>
> You might also want to check the block size that you're making your  
> file
> system with, that might give you some better I/O.
>
> -----Original Message-----
> From: Eric Yablonowitz [mailto:eyablon at tripadvisor.com]
> Sent: Monday, June 04, 2007 5:32 AM
> To: Vincent Janelle
> Cc: Peter Verhoeven; linux-poweredge at dell.com
> Subject: Re: MD1000 and PERC5/E I/O bottleneck
>
> full MD1000 w/ 15K RPM SAS HDDs.
>
> On Jun 4, 2007, at 12:22 AM, Vincent Janelle wrote:
>
>> Are these sata disks, or sas?
>>
>> -----Original Message-----
>> From: linux-poweredge-bounces at dell.com
>> [mailto:linux-poweredge-bounces at dell.com] On Behalf Of Eric
>> Yablonowitz
>> Sent: Sunday, June 03, 2007 6:37 PM
>> To: Peter Verhoeven
>> Cc: linux-poweredge at dell.com
>> Subject: Re: MD1000 and PERC5/E I/O bottleneck
>>
>> RAID-1 (inherently 2 disks) offers much less write throughput than a
>> 7 disk RAID-5 on the PERC-5 in my tests.  RAID-10 on 6 disks was
>> closer, but still (somewhat surprisingly) slower than RAID-5 on 7
>> disks.
>>
>> On Jun 3, 2007, at 9:21 PM, Peter Verhoeven wrote:
>>
>>> Can you afford to go with RAID-1 instead of RAID 5?
>>>
>>>
>>> -----Original Message-----
>>> From: linux-poweredge-bounces at dell.com
>>> [mailto:linux-poweredge-bounces at dell.com] On Behalf Of Eric
>>> Yablonowitz
>>> Sent: Saturday, 2 June 2007 9:03 AM
>>> To: linux-poweredge at dell.com
>>> Subject: MD1000 and PERC5/E I/O bottleneck
>>>
>>> So these results imply that there is a bottleneck somewhere that is
>>> preventing me from getting anywhere near what my 14 spindles can
>>> handle.
>>> Any thoughts on what this bottleneck might be?
>>>
>>> Thanks,
>>> Eric
>>>
>>
>
>
>