Re: no. of users supported

From: raja rao <raja4list_at_yahoo.com>
Date: Thu, 29 Sep 2005 01:15:19 -0700 (PDT)
Message-ID: <20050929081520.47584.qmail@web31406.mail.mud.yahoo.com>

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

99% of users are doing selects.

With this, can we have a guess, to howmany users this server supports.

(they wont stay idle, all the time they loggedin , they keep reading the things. this is the
nature of the db)

Can we guess the count now ?

26+26=52 is fine ?
or any good number near to that ?

Allan Nelson <anelson77388_at_gmail.com> wrote: There is not a hard and fast rule for situations like this. OLTP loads can vary widely. Consider for instance Oracle Financials which can be loosly characterized as an OLTP load over against a home grown application where say 52 people are just entering data from say a phone survey. Those would be very different load profiles.

Then there is the difference between people actively executing sql against your database and people who are just sitting there with a connection and not executing anything.

The best approach to seeing if this work might be a combination of watching the CPU and checking out what your database is waiting on. For a pure OLTP load and 2 CPU's I'd try and stay at or below about 70% of max cpu. If you go higher than that you may well get into queueing where response time for your users becomes unpredictable. Sorry not to give a more difinitive answer but it really isn't that cut and dried.

Allan

On 9/28/05, raja rao <raja4list_at_yahoo.com> wrote:Hi All,

os: Solaris 2.9

We have a solaris box with 2 GB RAM and 2 cpus.

Is there any rule to know how many oracle connections are fine to work with it (mean to say the oracle sessions on that).

I heard it is 26 users per cpu. is this true ? or is there any other principle to calculate how many users concurrently can use this box.

(the database is OLTP)

Thanks,
Raj

Yahoo! for Good
Click here to donate to the Hurricane Katrina relief effort.

--
http://www.freelists.org/webpage/oracle-l

Received on Thu Sep 29 2005 - 03:17:40 CDT