[prev in list] [next in list] [prev in thread] [next in thread] 

List:       openjdk-jdk6-dev
Subject:    Re: CMS output description
From:       Srinivas Ramakrishna <ysr1729 () gmail ! com>
Date:       2013-04-17 6:47:00
Message-ID: CABzyjyn8qnajyPA83GxHBhOorFiq+H1PBcppHhaahRAGJ1YWPg () mail ! gmail ! com
[Download RAW message or body]

Hi Venkat --

On Sun, Apr 14, 2013 at 12:43 PM,  <mndvenk@gmail.com> wrote:
> Ramki,
> 
> It is great to see quick response.
> 
> If the user time is too high comparatively to real time means is it short of \
> resources also what I have observed my application certainly stopped more than 5 \
> mins at the same time GC user time showing 3+ minutes.

The ratio of user:real is really a rough indicator of the level of
parallelism achieved by GC.
For the example you had below:-

> > Times: user=0.45 sys=0.00, real=0.09 secs

0.45/0.09 = 5 is the amount of parallelism in GC (of course not all of
that might be useful work, so
one can't with a straight face call it parallel speed-up, but still ...)

> 
> I will send cpu details with GC logs tomorrow.

OK.

-- ramki

> 
> Cheers,
> Venkat
> ------Original Message------
> From: Srinivas Ramakrishna
> To: Venkateswarlu Mandadi
> Cc: jdk6-dev@openjdk.java.net
> Subject: Re: CMS output description
> Sent: Apr 14, 2013 8:23 PM
> 
> Hi Venkat --
> 
> user: process virtual user time,
> sys: process-virtual system time
> real: elapsed time
> 
> Re: "Use showing more than 3 mins" : depends on how many cpus you have.
> 
> Really, what should typically concern you, as a user of GC, is the
> "real" time above, which is the duration for which your threads are
> stopped
> (when the time is associated with a stop-world pause). For concurrent
> ops, it's the end to end elapsed time for that cycle, so should not
> typically be of concern, as your application threads are running
> during that time.
> 
> cheers.
> -- ramki
> 
> 
> On Sat, Apr 13, 2013 at 1:37 PM, Venkateswarlu Mandadi
> <mndvenk@gmail.com> wrote:
> > Hi,
> > 
> > We are using CMS collector for oldgen (large heap -Xmx18g) and we have
> > observed there is a long whole world stop, in GC logs it was observed that
> > GC times are looking alright, but Times: Usr showing around more than 3
> > mins.
> > 
> > Can someone explain Times fields
> > 
> > Times: user=0.45 sys=0.00, real=0.09 secs
> > 
> > Best Regards,
> > 
> > Venkat
> 
> Sent via my BlackBerry from Vodacom - let your email find you!


[prev in list] [next in list] [prev in thread] [next in thread]