Hi there. Do not believe, what the systems requirements tell you. Oracle eats resources like popcorn. Can you give us an idea, what you are going to do with this server ? Rgds Alexander M. Ermes
.. and all these memories are going to vanish like tears in the rain! final words from Rutger Hauer in "Blade Runner"
Next questions : general usage, datawarehouse.... ? Go for Oracle RAC for high performance, if possible. Try to find out, how many users will really be connected to the system at the same time. I would suggest Proliant ML ( cluster ? ). Rgds Alexander M. Ermes
.. and all these memories are going to vanish like tears in the rain! final words from Rutger Hauer in "Blade Runner"
Consider first whether it will be used for OLTP or datawarehousing (business intelligence). You can see then what has been done to get high benchmark numbers at www.tpc.org. Look at the executive summaries or even more detail to understand some of what they did not only to get the processing power, but also the I/O that was needed. You may find your biggest challenge isn't in coupling multiple Proliant or Integrity servers, but in making sure that your shared storage system can handle the number of I/Os necessary. Make an estimate of how many I/Os per second you will need to support the 500 users and ask HP for a system than can deliver that number. Also, for some databases, Itanium will do much better due to the 64-bit implementation of Oracle. The best TPC-C result, last I checked, was a 10g cluster of 16 4-CPU Integrity (Itanium) rx5670 servers. You mention 500 users, but are they doing updates or queries. How complex are the queries? Are they all active users? What is the think time between updates or queries? Good luck.
I'm sitting watching 10g run idle on an HP-UX box, with no users at all and its top two processes are eating half the CPU cycles on a pretty powerful dual processer rp5450 server.
Worse than that, it doesn't seem to be balancing its load across the two CPU's very well.
I anticipate performance issues ahead.
SEP
Steven E Protter Owner of ISN Corporation http://isnamerica.com http://hpuxconsulting.com Sponsor: http://hpux.ws Twitter: http://twitter.com/hpuxlinux Founder http://newdatacloud.com
As reaction on your comments: I see a lot of dbsnmp CPU activity (constantly) on my HP servers. This is caused by the OMS (Oracle Management Server). Big processors are the only solution to that. The reason that sometimes one processor does nothing and the other one is "working it's ass off" is because Oracle assings one processor per thread. Cheers,
I'm only on the hp-ux, not oracle side of our systems, but on 11.11/oracle9 I also saw many issues regarding dbsnmp which by now have dissapeared, either due to version upgrade to 9.2.0.4 or some hotfix.
Maybe You can find something regarding that.
yesterday I stood at the edge. Today I'm one step ahead.
As a guide for you: We have a rp3410 with 6 GB and 1 800 MHz processor. Database has 50 GB and it´s installed in JBOD 36 GB disks 15k rpm. About 120 users and cpu hasn´t reached in 100%. Application is an ERP written in Forms. This UNIX box has 29000 tpm. Theoretically, you can use an DL380 with 3,4 GHz processor, but be carefull with Oracle prices. From 1 to 2 processors you can use Oracle Standard One, about US$ 5.000/processor, from 2-4 you can use Oracle Standard Edition, US$ 15.000/processor (for RAC it has to be this one) and for above, Oracle Enterprise Edition, US$ 40.000/processor. Sometimes it´s better to use a RISC box with less processors.
