Re: disk arms (was RE: Tips for user ASP)

[Vernon Hamberg <vhamberg@xxxxxxxxx>]

I think you're right. I just found this in the Performance Capabilities
manual, in the context of journaling and OptiConnect:

If RAID protection is used in the same environment, the DASD arms become
the bottleneck. If protection is required, using mirrored arms for the
journal receiver DASD will provide significantly better performance than
RAID protection. However, unprotected DASD arms will allow the highest
overall transfer rates and best overall performance when using Opticonnect
for remote journal.

OTOH, there is a chart showing, in a RAMP-C workload, that mirrored is
consistently faster than not mirrored.

At 12:57 AM 7/2/02 -0600, you wrote:
>On Mon, 1 Jul 2002, Vernon Hamberg wrote:
>
> > I was under the impression that there is some benefit from asynchronous
> > writes with mirrored sets. Which ever set is free gets the first write, the
> > other follows when it can. More queues to handle the requests.
>
>As I understand things the write is not complete until both sides of the
>mirror have been written to.  Otherwise you don't really have mirroring,
>do you?  But a read completes as soon as either side can fulfill the
>request.
>
>James Rich