on 02-05-2010 4:39 PM
We are planning a complete hardware refresh, and are considering two storage alternatives, both from IBM: DS8700 or XIV.
Currently:
We are running SAP Netweaver 7.00 & 7.01 on AIX 5.3 TL10 and Oracle 10.2.0.4
- The production ERP DB is about 6 TB and the production BI DB is about 2 TB.
- Weekday peak I/O per second are about 4,000 and 15,000 for our two current storage systems (DS8300 & DS8100, respectively). Weekend peak I/O per second are about 28,000 and 26,000.
- The production ERP system typically has up to 3000 peak concurrent users.
We have results of XIV capacity planning analyses compared to the DS8700. The DS8700 performs better, but it's not clear whether it's significantly better. It's the cost difference that makes the XIV a compelling option.
Does anyone have experience using IBM System Storage XIV in a similar (or any) SAP environment? What have you learned? Are any performance issues attributable to the storage system?
Hi,
Please refer this informative document "[Best Practices Guide for SAP Applications on an IBM XIV Storage System|http://www-03.ibm.com/systems/resources/storage_solutions_pdf_xiv_sap.pdf?CACHEID=4b3022004e454ceb91c6f3901a2417e7]" along with this informative [case study|ftp://ftp.software.ibm.com/common/ssi/pm/ab/n/spc03125wwen/SPC03125WWEN.PDF], to get more information.
Regards,
Bhavik G. Shroff
You must be a registered user to add a comment. If you've already registered, sign in. Otherwise, register and sign in.
Hi Dave,
We are running a landscape that is very similar to yours.
Netweaver 7.0 on AIX 5.3L TL 9 (11 on NIM and VIO) and Oracle 10.2.0.4
Our ERP DB is about 10TB and we are running a small CRM and BI landscape at about 2TB each.
We were previously splitting our landscape on EMC, ERP was on a DMX1000 and our advanced apps and QA env. were on a Clarrion CX380.
We have about 2500 concurrent users during peak hours.
We were an early adopter of XIV and so far it has exceeded our expectations. It has been very easy to maintain and after tuning we easily maintain all of our landscapes between about 6000 and 10,000 IOPS around regular peak hours. This obviously goes up around quarter close and other periods of intense usage.
My best advice is that the XiV itself has very little to configure, outside of either "round robin" or "failover" for your disk depending on your environment. Our sticking point was Oracle.
We had to intensively tune Oracle (no big surprise there) and adjust it's memory usage and other parms to really get significant performance out of the XiV. Once we did though, it levelled out any IO spikes and we got significant performance improvements. So I suggest either making your DB people aware that there will be changes needed, or if you do not have that resource maybe look into IBM providing some services on that front.
I also suggest making sure they install the latest microcode at setup time, because then all future microcode updates can be performed with no outages, which is very convenient.
Other than that things have gone well. We even lost a drawer at one point and experienced no downtime and only a minimal, short performance hit in our production landscape, so their claims about the grid architecture actually stood up to a real world test and performed well.
I hope this will help.
Thanks and good luck!
Edited by: Ben Daniels on Feb 17, 2010 9:32 AM
Edited by: Ben Daniels on Feb 17, 2010 9:38 AM
You must be a registered user to add a comment. If you've already registered, sign in. Otherwise, register and sign in.