> SCSI should do a better job for
> those parallel disk I/O than SATA, right?
This certainly used to be the case. However...
The newest SATA controllers and drives implement Native Command Queuing
(NCQ) which will improve performance in multiple-user environments. For
example, I believe the new Western Digital Raptor DW740GD (SATA
10,000rpm) outperforms many (much more costly) SCSI drives:
http://www.storagereview.com/articles/200401/20040126WD740GD_1.html
Thanks,
Jack
Xiong Jiang wrote:
> Hi,
>
> Great point, actually I only got the system recently (New Year gift :),
> thus I have not got a chance to run the benchmark testings yet. But
> personally there are two reasons I recommend SCSI over SATA:
> (1) I burned out one big external hard drive after intensive data
> analysis, thus I lost everything on it (although fortunatelly I backup
> it every day). I think SCSI should be more endurable than SATA, but I
> could be wrong.
> (2) Personally I tend to run two or more instances of SPM at the same
> time, a busy post-doc and computer :). SCSI should do a better job for
> those parallel disk I/O than SATA, right?
>
> But I would like to hear more from SPM experts (Matthew for sure :).
>
> Best,
>
> Xiong
>
>
> Matthew Brett wrote:
>
>> Hi,
>>
>>
>>
>>> (1) A SATA HD for opterating system;
>>> (2) A small but fast SCSI HD (~20-40G) for data processing;
>>>
>>>
>>
>> Just to say, that my guess would be that disk speed would not be a
>> large factor in SPM processing times, because even NFS over a 100Mb
>> link only had a minor effect compared to local hard disk:
>>
>> http://www.mrc-cbu.cam.ac.uk/Imaging/Common/spm_benchmarks.shtml
>>
>> I would be very interested to hear of any other data that people have
>> on this issue,
>>
>> Best,
>>
>> Matthew
>>
>>
>
>
|