RE: no. of users supported

  • From: "Kennedy, Jim" <jim_kennedy@xxxxxxxxxx>
  • To: <Nagarajan.Subbiah@xxxxxxxx>, "oracle list" <oracle-l@xxxxxxxxxxxxx>
  • Date: Thu, 29 Sep 2005 10:36:07 -0700

Too many variables.  I have seen just select environments only able to
run 1 user and on the same hardware, database etc. run 1,000 users. (2
CPUs 200 M Hertz.) It depends upon
        1. Use bind variables.
        2. Parse once, rebind and execute thereafter.
        3. Efficient SQL with the proper indexes etc.
        4. No or low row chaining.
        5. Array interface for fetches.
        6. Efficient and performant code.
        7. What the load is?  How many queries per unit time, how much
data?  Speed of disks, speed of processor, etc.

To know you really have to bench mark it with the type of load you are
going to use.  WAGs can really hurt.

Jim

(WAG -> Wild Assed Guesses)

-----Original Message-----
From: oracle-l-bounce@xxxxxxxxxxxxx
[mailto:oracle-l-bounce@xxxxxxxxxxxxx] On Behalf Of Subbiah, Nagarajan
Sent: Thursday, September 29, 2005 10:28 AM
To: 'oracle list'
Subject: RE: no. of users supported

It also depends on the processor speed. WE used to have issues with 8 -
400MHZ CPU with less than 100 connection and mostly select queries.
Assuming your hit ratio is close to 100% and with the average run time
of the all the queries are less than 20 milliseconds, I don't see any
issues with 2 - 1GHZ processor supporting at least 50 users.

-----Original Message-----
From: Dennis Williams [mailto:oracledba.williams@xxxxxxxxx]
Sent: Thursday, September 29, 2005 1:00 PM
To: raja4list@xxxxxxxxx
Cc: Allan Nelson; oracle list
Subject: Re: no. of users supported


Raj,

> The database is a complete read activity. not much inserts.

In general you are correct that only selects will reduce the server
load,
meaning the server should be able to support more users. But there are
several additional issues:
  - Are the queries just retrieving a few rows through indexes, or are
full-table scans required?
  - Do most of the queries go after the same data, so the popular block
are
cached in memory, or even use Materialized Views? Or will the queries
want
data all over the database? What is the I/O subsystem like? You may be
better off putting more money in the I/O subsystem than the server.

I have seen Solaris servers easily supporting the number of users you
have
mentioned.

Dennis Williams
--
//www.freelists.org/webpage/oracle-l
--
//www.freelists.org/webpage/oracle-l
--
//www.freelists.org/webpage/oracle-l

Other related posts: