[Memcacheddotnet-devel] RE: memcache exception

[Tim G. <ti...@ge...>]

I think I figured out what was wrong with the performance in the 1.1.* line
and I'm redeploying a new version.  Should be up in 10 minutes or so
(Sourceforge file release process is pretty cumbersome).

You can check the changelog or the news when it's up for a detailed
description of what was wrong, but my benchmarks go something like this now:

1.1.1: 10,000 gets/sets - 9.3s
1.1.2: 10,000 gets/sets - 8.2s

The other notable thing is that the CPU usage from 1.1.1 to 1.1.2 is down
from 30% to ~10%.

Lei and Max, I BCC'ed you and sent this email to the developer's mailing
list.

-Tim Gebhardt
ti...@ge...


-----Original Message-----
From: Lei Sun [mailto:le...@gm...] 
Sent: Tuesday, April 04, 2006 12:47 PM
To: ti...@ge...
Cc: Maxim Mass
Subject: Re: memcache exception

Hi Tim,

We tried to do a stress test on all 3 versions of the client.
Conditions:
1) 2 memcached server, both dual cpu, 4 GB memory
2) Client machine is p4 dual core 3.0G hyper threading
3) Basically a unit of our test is consisted of 100 set, 100, get, 100
getmulti, 1 flushall
4) We run a unit of test 300 times and 1000 times across all three
versions of the client.
Here is what we found:
	            1.0.3	1.1.0	 1.1.1
# of actions	sec to finish/exceptions		
30,000	         26		30
30,000	         25		30
30,000	         25		29
100,000	        81/0	  97/30	     101/0
100,000	        94/0	 109/47	    102/0
100,000	        82/0	 111/24	     97/0

Result:
1) 1.0.3 is actually the best in performance, network stats are steady
and high, and no exceptions
2) 1.1.0 is the worst in performance, network stats are zigzaggy and
lower, and lots of exceptions
3) 1.1.1's performance isn't that good, network stats are a little
better than 1.1.0, but still very zigzaggy comparing to 1.0.3. but no
exceptions

Good thing that you have reduced the exceptions from 1.1.1 comparing to
1.1.0!

Would you please look into this slower performance issue?

Thanks

Lei


On 3/31/06, Tim Gebhardt <ti...@ge...> wrote:
>
>
>
> The new release is up.  Let me know if you guys have any more problems.
>
>
>
> -Tim
>
>
>
>  ________________________________
>
>
> From: max...@gm... [mailto:max...@gm...] On Behalf Of Maxim
> Mass
>  Sent: Thursday, March 30, 2006 7:16 PM
>  To: ti...@ge...
>  Cc: le...@gm...
>  Subject: memcache exception
>
>
>
>
> We've been getting this exception pretty often in our prod environment.
> Seems to happen more as more load gets put on it. There's def. a bug with
> the client -- i dont think this is server related at all.
>
>  Here's a partial stack.. Unfortunately, i dont have the full trace.
>    at System.Net.Sockets.NetworkStream.Write(Byte[] buffer,
> Int32 offset, Int32 size)  at
> System.IO.BufferedStream.FlushWrite()  at
> System.IO.BufferedStream.Flush()  at System.IO.BufferedStream.Close()  at
> MemCached.clientlib.SockIO.TrueClose()  at MemCache
>
>
>  The exception.message is
>  Cannot access a disposed object named
> "System.Net.Sockets.NetworkStream".  Object name:
> "System.Net.Sockets.NetworkStream".
>
>
>  Have you been experiencing anything like this? can you tell me the
settings
> you've been using for the client?
>  We're still using previous version right before your fxcop changes. Do
you
> know if that version fixes anything?
>
>  Max
>
>