Re: AIX and Oracle

Joseph Miller wrote:
>
> I am on a project to set up a database of 250,000 people. I think I will
> end up using two servers and I am trying to figure out if I should share
> a single database, or if I should mirror a database. It will probably be
> an RS/6000 smp cluster with Aix 4.***.
>
> Any info will be helpful.
>
> JMMiller_at_OSC.State.ny.us

You are talking about "I should mirror a database", do you mean replication? If so, the last time I looked at replication it could only handle 20 tps (transcations per second). Snapshots can handle more if you need the power.

You also talked about "share a single database" and "RS/6000 smp cluster". I assume you are talking about the IBM SP2 environment. The remainder of this mail is from a previous posting of mine. It's intention is only to make people think before they leap:

---START OF OLD POSTING---
I posted these links in this news group a few days ago. If you take the time to realy look/read everything here you will find things from tuning, setup, staffing, how it works, etc. You will also read a few things you will not like. But thats the same with anything.

Try http://www.sequent.com/public/news/papers.html
    http://www.sequent.com/public/news/papers/ops/index.html
    http://www.sequent.com/public/news/papers/smp_mpp/index.html

Several interesting quotes from the papers are:

Users requiring high-performance UNIX database systems should fully exploit SMP -- with its mature software library -- before considering MPP systems. Most users will never use up the increasing power of SMP systems. Those that eventually do exhaust SMP will benefit by having waited for MPP systems to attract more native software and better systems and data management tools.

Leading RDBMSs (e.g., Oracle7 and Informix Dynamic Server®) function better on SMP systems than MPP systems. Oracle Parallel Query runs on MPP systems but requires the hardware vendor file system to simulate a shared-disk system. This requires the file system and database to use the interconnect for I/O shipping -- shipping all the data to the executing node -- instead of function shipping. The result can be extensive data traffic, latency and degraded performance.

While both SMP and MPP systems can [Process complex queries], the difficulties inherent in setting up and tuning an MPP database can outweigh any performance benefit.

Lack of management tools for MPP systems, either from MPP vendors or third parties, is a major problem. Most MPP-based OLTP and DSS systems successes are the result of long, hard programming efforts where an organization can tolerate higher costs, longer delays and greater uncertainty.

A recent survey found no "dissatisfied" SMP users, but found 8 percent of MPP users to be "somewhat dissatisfied" and 15 percent "very dissatisfied."     

--
                       \\|//
                       (0-0)
           +-----oOO----(_)-----------+
           | Brian P. Mac Lean        |
           | Database Analyst         |
           | brian.maclean_at_teldta.com |
           | http://www.teldta.com    |
           +-------------------oOO----+
                      |__|__|
                       || ||
                      ooO Ooo