Re: R.A.I.D boxes

In <CzDLv2.F7t_at_world.std.com> mew_at_world.std.com (Michael E Willett) writes:

>
>
>>>RAID 3 fails on small reads and small writes and RAID 5 handles
>>>large and small writes poorly, while RAID 7 consistently outperforms
>>>the single spindle rate. The RAID 7 performance multiple increase
 over
>>>RAID 3 and RAID 5 is a function of the small read/small write/large
>>>mix.
>
>Daniel Jones writes:
>
>>Michael's statement is an exercise in semantics since his unspoken
>>assertion is that the assumed deficiencies of RAID 3 and RAID 5 are
>>based on uncached implementations....
>
>I have independently emailed Daniel some architectural RAID information
>that should clear this up. There's much more to the big improvement in
>RAID 7 performance over the earlier, mid-80s RAID architectures than
>just the advanced caching techniques.

Thanks, Mike, but you still seem to be confused about terminology. Specifically, the two words architecture and implementation are not interchangeable. I am sure that your product can do well against specific RAID 3 and RAID 5 subsytems. However, a generic comparison of the performance of RAID 7 to RAID 3 or RAID 5 is a content null statement. RAID 7 is neither faster or slower. Received on Wed Nov 16 1994 - 22:49:17 CET