On 12/11/04 5:05 am, in article
a2efdbbd.0411112105.d210392@posting.google.com, "Dean" <dean.sun@gmail.com>
wrote:

> I did a write IO test in IBM P690 and HP SUPERDOME.
> When I tested the local disks in P690,use this:
> ----------------------------------------------------
> time dd if=/dev/zero of=/tmp/test.dat bs=1024 count=300000
> 300000+0 records in
> 300000+0 records out
> real 0m42.01s
> user 0m0.85s

> I'll cry for explaining this to the customer.
> Anyone can help me to explain this?

I'm sure you are not trying to sell the benefits of running dd to the
customer. Performance comparisons are very hard to quantify properly unless
you install an exact duplicate of the app / database on both systems and
measure actual throughput by simulating the customers expected experience -
expensive and hard to do and even then, tuning would probably improve
performance on either system.

Unfortunately this usually comes down to the preference and comfort factor
of the customer i.e. If they are used to running HP, then stick with HP.
Same with IBM unless there is an overwhelming reason to use either i.e. You
want to use LPARing specifically on IBM / you have more HP people than IBM /
you need high availability rather than absolute performance etc. From
experience of HP / Sun and IBM, these systems can all be coaxed into giving
adequate performance to customer requirements unless you are vying for the
Super500 computer award then obviously go for IBM ;0)