| Load on Oracle server [message #219215] |
Tue, 13 February 2007 10:29  |
manish mendiratta
Messages: 110
Registered: May 2002
|
Senior Member |
|
|
Hello All,
We have an OLTP database which is accessed by > 200 Concurrent users at a time in MTS mode(we are a dotcom company). I use uptime command to get load on server.
$ uptime
10:56am up 145 day(s), 16:04, 4 users, load average: 4.00, 4.85, 5.29
My question is what should be the ideal load for such servers? I have read some where that if load goes beyond 1 then there's a problem.
If this is a problem how can i have it fixed or what are the other methods thru which i can confirm this is a problem?
Please shed some light on this from your environment experiences.
Thanks
Manish
|
|
|
|
| Re: Load on Oracle server [message #219251 is a reply to message #219215] |
Tue, 13 February 2007 12:30  |
smartin
Messages: 1803
Registered: March 2005
Location: Jacksonville, Florida
|
Senior Member |
|
|
This is actually more of an operating system specific question than Oracle, but they are certainly related. But you may have more luck on Unix OS specific websites and forums is my point.
I believe those load numbers would only be relative in the context of the number of processors you have on your server.
In general, check the man pages for the commands you are issuing, and they should tell you exactly what each value reported means.
There are other performance and server measurement commands to use, such as sar and iostat. Unix is quite full of useful commands such as these, but many behave slightly differently depending on the exact unix or linux version you are on.
But just as importantly, if you are a large .com site with lots of concurrent users, you should have a dedicated system administrator to take the lead in these matters, with you as the dba in theory taking the lead of database matters.
And the two of you working together to identify and resolve issues. If your company is large, and doesn't have a person in each role, and thinks that it can get by with just having the DBA also do unix sys admin on the side, then I'd view this as a very large mistake.
|
|
|
|
| Re: Load on Oracle server [message #219259 is a reply to message #219215] |
Tue, 13 February 2007 12:59 |
manish mendiratta
Messages: 110
Registered: May 2002
|
Senior Member |
|
|
Thanks for your reply smartin,
Well yes i know it's a sys admin job, but i have been asked this question so i have to reply.
i know uptime means the avg number of jobs in the run queue.
But could this be a CPU bottle neck?
I have used IOSTAT and VMSTAT to co-relate data but have found anything so far. What's your experience , I mean load on your
server etc?
Uptime clearly doesnot states that it's waiting for something. It's just that it's has been running so many jobs at that time.
I am not sure if my this interpretation is correct.
Thanks
Manish
|
|
|
|
| Re: Load on Oracle server [message #219260 is a reply to message #219215] |
Tue, 13 February 2007 13:16 |
 |
BlackSwan
Messages: 26766
Registered: January 2009
Location: SoCal
|
Senior Member |
|
|
Looking at only 1 metric is borderline useless exercise.
top - 11:12:45 up 66 days, 16:52, 4 users, load average: 14.59, 12.14, 12.26
Tasks: 284 total, 2 running, 262 sleeping, 0 stopped, 20 zombie
Cpu(s): 7.6% us, 2.7% sy, 0.0% ni, 76.4% id, 13.0% wa, 0.0% hi, 0.2% si
Mem: 12182932k total, 12162940k used, 19992k free, 12304k buffers
Swap: 13197388k total, 61680k used, 13135708k free, 9365840k cached
This system has "high" run queue size, but is about 75% idle while IO Wait is at 13%
So which is the best measure of system "load"; run queue or %idle?
|
|
|
|
] 
Blog
Search
Help
Members
Register
Login
Home
