Now Im puzzled, 
I wasnt aware of the fact you have to pay for an PAV license.
I didnt read the whole "story".

I was sure, that PAV is only an eckd-linux-driver bussiness, wich isnt 
developed/released yet.

Perf: The goal is, to get the most perf, lowest response time, best scalabilty, and of 
course the most stable out of that, what is given to you.
Some special features may be suitable for special environments, but unfortunately many 
salespersons an buyers calculating cost/performance ratio exactly like its named. 
Cost/Performance... 

but lets deal again with the subject: Maybe it would help too, when you are able to 
get the system disks, (v)swap disks (also Swap of VM) an the ext3-journal of your big 
fs away from the certainly heavy utilized CHPIDs. Stripe as much as you can. logical 
and physical. You loose the flexibility of growing...

I think the performance of an M9 disk isnt important. 
We benchmarked striped M3 and M9 pvs, same LVM-layout, had 4 and 8 phys. 
escon-channels. Nearly Same perf-values.
Aren't the DASDs virtual in the storage boxes?? Tracks are living anywhere!? 


regards
fs

> Frank Schwede
> Lufthansa Systems Infratec GmbH
> mailto:[EMAIL PROTECTED]

-----Urspr�ngliche Nachricht-----
Von: Linux on 390 Port [mailto:[EMAIL PROTECTED] Auftrag von
Rob van der Heij
Gesendet: Dienstag, 20. Juli 2004 09:09
An: [EMAIL PROTECTED]
Betreff: Re: WG: Performance of large file systems


Frank Schwede, LSY wrote:

>Im sorry, but:
>What is going on here?
>
>I was sure, someone would give a statement about the real current state of PAV on 
>Escon DASDs.
>
>
Maybe others were puzzled like myself. Why the &$* would you pay for PAV
license on the ESS and not invest in FICON?  When all response times are
cut so drastically, the trouble of waiting in queue is less dramatic.

If you need to decide between small disks and large ones with PAV, can
think of scenario's where the PAV solution is better. With such large
filesystems you already took the cost of LVM anyway. The benefit of
striping does depend a lot on application behaviour, and it is easy to
make it counter productive. I am having a hard time trying to come up
with proper benchmarks to compare scenario's. This is complicated and
there are so many variables, each answer that I find gives me two more
questions.

Rob

----------------------------------------------------------------------
For LINUX-390 subscribe / signoff / archive access instructions,
send email to [EMAIL PROTECTED] with the message: INFO LINUX-390 or visit
http://www.marist.edu/htbin/wlvindex?LINUX-390

----------------------------------------------------------------------
For LINUX-390 subscribe / signoff / archive access instructions,
send email to [EMAIL PROTECTED] with the message: INFO LINUX-390 or visit
http://www.marist.edu/htbin/wlvindex?LINUX-390

Reply via email to