IDS Forum

Re: IDS instance memory

Posted By: Art Kagel
Date: Friday, 17 July 2009, at 11:03 a.m.

In Response To: Re: IDS instance memory (KAMRAN HAQ)

An IDS engine instance consists of several oninit processes, known
internally as Virtual Processors or VPs, performing different functions for
the engine. If you run onstat -g glo you will see all of the oninit
processes (VPs) and what VP class each is running as. At a minimum you will
have:

- one or more CPU VPs - these process user sessions and may poll for

connections depending on configuration. The CPU VPs also maintain a second

OS level light weight thread of execution known as the KAIO thread that

manages IO and communicates with the OS Asynchronous IO mechanism.

- one ADM or administration VP - Performs some housekeeping functions and

oversees the health of the other VPs.

- one MSC or miscellaneous VP - another housekeeping virtual processor

- one or more AIO VPs - these handle asynchronous IO to file types that

do not support OS level asynchronous IO like COOKED file including COOKED

chunks that are not opened O_DIRECT and text files like the message log

(onstat -m).

- zero or more NET VPs (soc, tli, str, etc.) - these can be configured to

handle polling for connection and work requests from users to offload some

of that from the CPU VPs where appropriate.

- User VPs - virtual processors created by you to execute User Defined

Routines (UDRs) that may impact the server's performance or which are not

yet completely tested. This isolates UDRs from the mainline server

processes to prevent a UDR from crashing the server.

There are others, but these are the main ones. As to the total memory,
almost all of that memory is shared including the oninit code space. If you
run 'size $INFORMIXDIR/bin/oninit' the last one or two values listed (see
the size man page) are usually the executable's private memory. This is the
only memory that is not shared between the oninits and so those blocks are
the only memory you can add up across the number of oninit processes. The
remainder only counts once as does the shared memory.

You can use the output below to estimate this. Take the 926MB once and add
in the difference between that and the value in the next column for each
oninit. So below, of these are the only oninit processes, the total use is
close to (but actually a bit lower than): 926 + 116 + 57 + 120 + 86 =
~1305MB

Art S. Kagel
Oninit (www.oninit.com)
IIUG Board of Directors (art@iiug.org)

Disclaimer: Please keep in mind that my own opinions are my own opinions and
do not reflect on my employer, Oninit, the IIUG, nor any other organization
with which I am associated either explicitly or implicitly. Neither do
those opinions reflect those of other individuals affiliated with any entity
with which I am affiliated nor those of the entities themselves.

On Fri, Jul 17, 2009 at 10:24 AM, KAMRAN HAQ <khaq@i2cinc.com> wrote:

> When virtual segment of memory are added then shared memory appears
> increased
> by that segment, is there any other memory overhead on addition of virtual
> segment?
> Also why there appears more than one occurences for single running IDS
> instance in prstat output. If we add these numbers to find memory consumed,
> it
> leads to far large compared to actual used memory. as following prstat
> output
> shows around 4 GB consumed by oninit process where "oninit -" shows it
> around
> 1 GB
> ---------------------------------------------------------------------
>
> 8017 informix 926M 810M sleep 50 -10 0:21:22 2.0% oninit/1
> 8009 informix 926M 869M cpu2 59 -10 1:07:06 2.0% oninit/1
> 8016 informix 926M 814M sleep 50 -10 0:26:01 1.9% oninit/1
> 8011 informix 926M 840M cpu3 59 -10 1:08:27 1.9% oninit/1
>
>
>
> *******************************************************************************
> Forum Note: Use "Reply" to post a response in the discussion forum.
>
>

--001636c5afb6332abf046ee818e5

Messages In This Thread

• IDS instance memory
  KAMRAN HAQ -- Friday, 17 July 2009, at 8:51 a.m.
  • Re: IDS instance memory
    KAMRAN HAQ -- Friday, 17 July 2009, at 9:35 a.m.
    • Re: IDS instance memory
      Art Kagel -- Friday, 17 July 2009, at 10:42 a.m.
  • Re: IDS instance memory
    Art Kagel -- Friday, 17 July 2009, at 9:48 a.m.
    • Re: IDS instance memory
      KAMRAN HAQ -- Friday, 17 July 2009, at 10:24 a.m.
      • Re: IDS instance memory
        Art Kagel -- Friday, 17 July 2009, at 11:03 a.m.

IDS Forum is maintained by Administrator with WebBBS 5.12.