Hey everyone,

What type of garbage collection patterns are you seeing on your production
systems?  Has anyone experimented and determined what the optimal pattern
should be?  We are seeing major GCs about every 6 minutes.. during that time
the db connections spike, response times get larger, etc..  We have
configured the servers so that they use as much memory as possible (before
paging) but maybe performance (as a whole) would be better to configure a
smaller amount of memory with more frequent GCs?  We currently allocate
~1.7G of RAM on each app server and have about 70 req/sec load (as reported
by the web container pool).  Anyone have any data to share?

-Matt

[ Post a follow-up to this message ]

As a rule of thumb, the server should not spend more than 15% of the time
doing GC. The calculation is as follows:

% time spent in GC = (Avg. GC duration) / (Avg. GC duration + Avg. interval
between GC)

Try increasing the -Xms value to make GCs less frequent.

--
Thank you,

Bibhas Bhattacharya
--
Web Age Solutions Inc.
WebSphere and WebLogic Training and Consulting Services
www.webagesolutions.com

"Matt Fleming" <xagloel02@sneakemail.com> wrote in message
news:brntj3$640c$1@news.boulder.ibm.com...
quote:

> Hey everyone,
>
> What type of garbage collection patterns are you seeing on your production
> systems?  Has anyone experimented and determined what the optimal pattern
> should be?  We are seeing major GCs about every 6 minutes.. during that

time
quote:

> the db connections spike, response times get larger, etc..  We have
> configured the servers so that they use as much memory as possible (before
> paging) but maybe performance (as a whole) would be better to configure a
> smaller amount of memory with more frequent GCs?  We currently allocate
> ~1.7G of RAM on each app server and have about 70 req/sec load (as

reported
quote:

> by the web container pool).  Anyone have any data to share?
>
> -Matt
>
>

[ Post a follow-up to this message ]

Yeah we have the initial and the max heap sizes set the same.  15% based
upon your calc seems pretty high.. we are nowhere near that level of GC.

-Matt
"Bibhas Bhattacharya" <bibhas@webagesolutions-NO-SPAM.com> wrote in message
news:brstej$64vk$1@news.boulder.ibm.com...
quote:

> As a rule of thumb, the server should not spend more than 15% of the time
> doing GC. The calculation is as follows:
>
> % time spent in GC = (Avg. GC duration) / (Avg. GC duration + Avg.

interval
quote:

> between GC)
>
> Try increasing the -Xms value to make GCs less frequent.
>
> --
> Thank you,
>
> Bibhas Bhattacharya
> --
> Web Age Solutions Inc.
> WebSphere and WebLogic Training and Consulting Services
> www.webagesolutions.com
>
> "Matt Fleming" <xagloel02@sneakemail.com> wrote in message
> news:brntj3$640c$1@news.boulder.ibm.com... 
production[QUOTE] 
pattern[QUOTE] 
> time 
(before[QUOTE] 
a[QUOTE] 
> reported 
>
>

[ Post a follow-up to this message ]

> Yeah we have the initial and the max heap sizes set the same.

What is is your -Xms value? To reduce GC frequency increase the -Xms value.
Setting -Xms same as -Xmx simply eliminates heap growth.


--
Thank you,

Bibhas Bhattacharya
--
Web Age Solutions Inc.
WebSphere and WebLogic Training and Consulting Services
www.webagesolutions.com

"Matt Fleming" <xagloel02@sneakemail.com> wrote in message
news:brt3ik$65f4$1@news.boulder.ibm.com...
quote:

> Yeah we have the initial and the max heap sizes set the same.  15% based
> upon your calc seems pretty high.. we are nowhere near that level of GC.
>
> -Matt
> "Bibhas Bhattacharya" <bibhas@webagesolutions-NO-SPAM.com> wrote in

message
quote:

> news:brstej$64vk$1@news.boulder.ibm.com... 
time[QUOTE] 
> interval 
> production 
> pattern 
that[QUOTE] 
> (before 
configure[QUOTE]
> a 
allocate[QUOTE] 
>
>

[ Post a follow-up to this message ]

Hi Matt,

You say that you have the min and max heap sizes the same (1.7 gig). That's
where you problem is. The garbage collector only kicks in when you have no
more heap memory available to allocate. When it does run in your case it has
a lot of work to do because it probably is collecting over  1 gig of
garbage.

By reducing -Xms the heap will only grow to the size it needs (up
from -Xms). Garbage collection will be more frequent but of a shorter
duration.

Cheers,
Glen.

"Matt Fleming" <xagloel02@sneakemail.com> wrote in message
news:brt3ik$65f4$1@news.boulder.ibm.com...
quote:

> Yeah we have the initial and the max heap sizes set the same.  15% based
> upon your calc seems pretty high.. we are nowhere near that level of GC.
>
> -Matt
> "Bibhas Bhattacharya" <bibhas@webagesolutions-NO-SPAM.com> wrote in

message
quote:

> news:brstej$64vk$1@news.boulder.ibm.com... 
time[QUOTE] 
> interval 
> production 
> pattern 
that[QUOTE] 
> (before 
configure[QUOTE]
> a 
allocate[QUOTE] 
>
>

[ Post a follow-up to this message ]