[Search for users] [Overall Top Noters] [List of all Conferences] [Download this site]

Conference ssag::ask_ssag

Title:Ask the Storage Architecture Group
Notice:Check out our web page at http://www-starch.shr.dec.com
Moderator:SSAG::TERZAN
Created:Wed Oct 15 1986
Last Modified:Fri Jun 06 1997
Last Successful Update:Fri Jun 06 1997
Number of topics:6756
Total number of notes:25276

6454.0. "HSZ Config Quest" by EPS::CONNORS () Thu Mar 06 1997 16:40

    (cross posted in hsz40 conference)
    
We are in the process of developing a sizing guide
for one of our partners suite of applications.  The I/O
section is a real challenge.  Some end-user configurations
will require recommending large amounts of storage.  We'd
like to understand the realistic break point for a single
PCI slot .... 

4100 Example

Is it wise to fully populate the 5 available PCI slots
with KZPSA's ?
 
Should we limit one HSZ50 per KZPSA ?  Is it unwise 
to daisy chain two, three, etc. ?

Should we limit, for example, 24 disks - 6 per channel behind
the KZPSA ?

We are very concerned with performance.  The environment will
include PeopleSoft on Oracle.  The sizing guide must include 
recommendations at various "user" levels along with various 
Production, QA, Development and Test Database size's.   
Some systems will contain all
of the PeopleSoft modules, some only a few.  

We understand the best way to layout the filesystems, ex;
roll back segs, redo logs, Oracle System Table Space, etc. We'll
have multiple database instances on the server(s) being sized. 

We definately have the strong potential to max out the I/O 
capabilities of a 4100 and need to know when to jump to the 
next level. (Ex: second server or 8400).

Any wisdom/insight would be appreciated.  

Thanks in advance
T.RTitleUserPersonal
Name
DateLines
6454.1MSE1::PCOTEpress one now for personal nameWed Jun 04 1997 09:4130
Good note as I have similar concerns.

>Is it wise to fully populate the 5 available PCI slots
>with KZPSA's ?

 Whether the 4100 can can sustain that kind of I/O is one
 question but whether you want to put ALL THAT DATA on
 just one server (single point of failure) is probably
 more of a concern.
 
>Should we limit one HSZ50 per KZPSA ?  Is it unwise 
>to daisy chain two, three, etc. ?

 One of the StorageWorks engineers mentioned that more
 then 2 HSZ50s "throttles" a single SCSI bus. (his words). Note, 
 that did not imply a dual-redundant configuration, it was
 two seperate disk arrays as in two SW450s.

 In addition, putting more than two hsz50s worth of storage 
 (We're talking close to a terabyte now with rz40s) on one
 bus may not be prudent since the SCSI bus is a single 
 point of failure.

 Two more considerations. Devices supported by the hsz50
 are fast10/narrow. Devices supported by the hsz70 can negotiate
 ultra/wide. A significant performance increase.

 ALso, have you considered SCSI based clusters ( VMS,ASE,Wolfpack)
 for scablability / high availability reasons?