cancel
Showing results for 
Search instead for 
Did you mean: 

IDLE TIME is more

IDLE TIME is more

HI ALL

Recently i checked my linux machine with IOSTAT command. Result is

avg-cpu: %user(4.49),nice(0.00) %sys(2.03)
%iowait(36)

what might be reason like ? what to do reduce IOWAITs?
7 REPLIES
Gerardo Arceri
Trusted Contributor

Re: IDLE TIME is more

Hard to tell without knowing the purpose of the server, but high i/o wait times normally indicates a bottleneck on the I/O subsystem.
What's running on the server ?
TwoProc
Honored Contributor

Re: IDLE TIME is more

To reduce I/O waits(in general, with no other data to make decisions with).

A) do less I/O (tuning code or run code at different times to load-level).
B) have more hardware capacity to do more I/O with more controllers, more disks, faster array subsystems, etc.
We are the people our parents warned us about --Jimmy Buffett

Re: IDLE TIME is more

Hi,

>>what might be reason like ?
There is a IO bottleneck
What is the IO response time on your system?

>>what to do reduce IOWAITs?
You need to identify the bottleneck. Is your IO's directed to your local disk, storage array, EVA or XP ?

Bottlenecks can be the disk itself or the disk controller.

If response time to your disk not good then if it is a local disk you may need to replace it or if it is a SAN you may need think about reconfiguring it.

-Pramod.

Re: IDLE TIME is more

What are the parameters i can check for this issue ? Server is running oracle DB. When i do TOP command its showing 4 oracle processes.

Re: IDLE TIME is more

Hi Promod

its direct server only.

Re: IDLE TIME is more

HI Madhu

Please find ouput..
TwoProc
Honored Contributor

Re: IDLE TIME is more

If it's running Oracle - you need to get your DBA to run a statspack and explain his highest activity in each section, and then he must work to tune those pieces of code. He'll need to refine code by working with development teams, create indexes, schedule gathers, work with 3rd party vendors to streamline processes. He also needs work with you and review setup of shared memory spaces, buffer caches, and operating system file I/O cache areas. He/she needs to examine what code out there is experiencing waits, and find ways for that code to have less impact on system for tuning. Also, have that person review their cache hit ratios and total I/O - which can sometimes be reduced with bigger buffer_cache areas, depending on lots of variables (initrans, free space allocation, redo log interleaving, temp spaces, code caching, etc). It's very involved, make sure he/she has good experience or at least good reference materials and tools. Methinks your team is probably light in the DBA area though, otherwise you probably wouldn't be asking these questions.
We are the people our parents warned us about --Jimmy Buffett