You are viewing a plain text version of this content. The canonical link for it is here.
Posted to users@mina.apache.org by Peter Frohberg <em...@pre-dev.de> on 2011/08/03 16:41:22 UTC
Vysper performance analysis
Hi all,
I'm a computer scientist student at the German university TU Dresden.
The past six months I wrote my minor thesis focusing on the scalibility of Vysper.
Due the fact that the results are written in German, I tried to write a short summary in English, which I'd like to share and can be found here:
http://www.pre-dev.de/vysper/SummaryAnalysisVysper2011.pdf
If you have any questions redarding this summary, feel free to ask and I try my best to anwser them as detailed as possible.
Greetings,
Peter
Re: Re: Re: Re: Vysper performance analysis
Posted by Peter Frohberg <em...@pre-dev.de>.
7 physical senders (6 virtual senders per physical sender) -> 42 virtual senders
25 physical receivers (20 virtual receivers per physical receiver) -> 500 virtual receivers
The reason for this low number of virtual senders per physical sender was the memory management of my test clients. They had to write all sent messages to the database having low thread priority.
> If I read things correctly, that would have been with 14*6 senders and
> 14*20 receivers?
>
> /niklas
>
Re: Re: Re: Vysper performance analysis
Posted by Niklas Gustavsson <ni...@protocol7.com>.
On Thu, Aug 4, 2011 at 8:19 PM, Peter Frohberg <em...@pre-dev.de> wrote:
> And to anwser your first question: When starting the test, all clients sent and received the messages in the way it was supposed. After 30 - 60 seconds most of the clients disconnected and just a few where able so send further messages - but no client received any of those messages. And yes - had to restart the service.
If I read things correctly, that would have been with 14*6 senders and
14*20 receivers?
/niklas
Re: Re: Re: Vysper performance analysis
Posted by Peter Frohberg <em...@pre-dev.de>.
Hello David,
no, the test environment doesn't exist any more. I just "borrowed" parts of the universities network infratructure for a limited time.
And to anwser your first question: When starting the test, all clients sent and received the messages in the way it was supposed. After 30 - 60 seconds most of the clients disconnected and just a few where able so send further messages - but no client received any of those messages. And yes - had to restart the service.
Greetings,
Peter
----- Original Message -----
From: dvlato@gmail.com
To: users@mina.apache.org
Date: 04.08.2011 12:26:29
Subject: Re: Re: Vysper performance analysis
> Hello Peter,
>
> It's not clear to me what your "error" was. I mean, did the clients got
> disconnected? In the past, some users mentioned that after some time
> working, their server 'hang' with lots of Broken Pipe exceptions and that
> their clients connected but never received a response... Is that what
> happened to you? Do you need to 'reboot' vysper in order to process new
> messages?
>
>
> If you still have your test environment ready and it's not much work for
> you, it would be very useful that you could get a thread dump ( if you also
> add a memory dump, that would be great) from any moment after vysper stops
> working.
>
>
> Cheers,
>
> David
>
>
>
> 2011/8/3 Peter Frohberg
>
> > Hi Niklas,
> >
> > I'm sorry but I really don't have any details that would help you. I did
> > not try logging on debug level becaus the first 'test'-tests I ran Vysper
> > with default (debug) logging I got the following error two times - before
> > figuring out, that I should change the logging level:
> >
> > java.net.SocketException: Broken pipe
> > at java.net.SocketOutputStream.socketWrite0(Native Method)
> > at
> > java.net.SocketOutputStream.socketWrite(SocketOutputStream.java:92)
> > at java.net.SocketOutputStream.write(SocketOutputStream.java:136)
> > at
> > com.sun.net.ssl.internal.ssl.OutputRecord.writeBuffer(OutputRecord.java:297)
> > at
> > com.sun.net.ssl.internal.ssl.OutputRecord.write(OutputRecord.java:286)
> > at
> > com.sun.net.ssl.internal.ssl.SSLSocketImpl.writeRecordInternal(SSLSocketImpl.java:743)
> > at
> > com.sun.net.ssl.internal.ssl.SSLSocketImpl.writeRecord(SSLSocketImpl.java:731)
> > at
> > com.sun.net.ssl.internal.ssl.AppOutputStream.write(AppOutputStream.java:59)
> > at sun.nio.cs.StreamEncoder.writeBytes(StreamEncoder.java:202)
> > at sun.nio.cs.StreamEncoder.implFlushBuffer(StreamEncoder.java:272)
> > at sun.nio.cs.StreamEncoder.implFlush(StreamEncoder.java:276)
> > at sun.nio.cs.StreamEncoder.flush(StreamEncoder.java:122)
> > at java.io.OutputStreamWriter.flush(OutputStreamWriter.java:212)
> > at java.io.BufferedWriter.flush(BufferedWriter.java:236)
> > at
> > org.jivesoftware.smack.PacketWriter.writePackets(PacketWriter.java:193)
> > at
> > org.jivesoftware.smack.PacketWriter.access$000(PacketWriter.java:40)
> > at org.jivesoftware.smack.PacketWriter$1.run(PacketWriter.java:76)
> >
> > (but maybe this already helps you. ;-) )
> >
> > Greetings,
> > Peter
> >
> > ----- Original Message -----
> > From: niklas@protocol7.com
> > To: users@mina.apache.org
> > Date: 03.08.2011 16:54:24
> > Subject: Re: Vysper performance analysis
> >
> >
> > > Very interesting, thanks for notifying us! I'm sure the german version
> > > would be of interest as well, there are people around here who would
> > > have no problem understanding it :-)
> > >
> > > Would you happen to have further details on the errors you got with
> > > the high frequency test cases?
> > >
> > > /niklas
> >
> >
> >
> >
>
Re: Re: Vysper performance analysis
Posted by David Latorre <dv...@gmail.com>.
Hello Peter,
It's not clear to me what your "error" was. I mean, did the clients got
disconnected? In the past, some users mentioned that after some time
working, their server 'hang' with lots of Broken Pipe exceptions and that
their clients connected but never received a response... Is that what
happened to you? Do you need to 'reboot' vysper in order to process new
messages?
If you still have your test environment ready and it's not much work for
you, it would be very useful that you could get a thread dump ( if you also
add a memory dump, that would be great) from any moment after vysper stops
working.
Cheers,
David
2011/8/3 Peter Frohberg <em...@pre-dev.de>
> Hi Niklas,
>
> I'm sorry but I really don't have any details that would help you. I did
> not try logging on debug level becaus the first 'test'-tests I ran Vysper
> with default (debug) logging I got the following error two times - before
> figuring out, that I should change the logging level:
>
> java.net.SocketException: Broken pipe
> at java.net.SocketOutputStream.socketWrite0(Native Method)
> at
> java.net.SocketOutputStream.socketWrite(SocketOutputStream.java:92)
> at java.net.SocketOutputStream.write(SocketOutputStream.java:136)
> at
> com.sun.net.ssl.internal.ssl.OutputRecord.writeBuffer(OutputRecord.java:297)
> at
> com.sun.net.ssl.internal.ssl.OutputRecord.write(OutputRecord.java:286)
> at
> com.sun.net.ssl.internal.ssl.SSLSocketImpl.writeRecordInternal(SSLSocketImpl.java:743)
> at
> com.sun.net.ssl.internal.ssl.SSLSocketImpl.writeRecord(SSLSocketImpl.java:731)
> at
> com.sun.net.ssl.internal.ssl.AppOutputStream.write(AppOutputStream.java:59)
> at sun.nio.cs.StreamEncoder.writeBytes(StreamEncoder.java:202)
> at sun.nio.cs.StreamEncoder.implFlushBuffer(StreamEncoder.java:272)
> at sun.nio.cs.StreamEncoder.implFlush(StreamEncoder.java:276)
> at sun.nio.cs.StreamEncoder.flush(StreamEncoder.java:122)
> at java.io.OutputStreamWriter.flush(OutputStreamWriter.java:212)
> at java.io.BufferedWriter.flush(BufferedWriter.java:236)
> at
> org.jivesoftware.smack.PacketWriter.writePackets(PacketWriter.java:193)
> at
> org.jivesoftware.smack.PacketWriter.access$000(PacketWriter.java:40)
> at org.jivesoftware.smack.PacketWriter$1.run(PacketWriter.java:76)
>
> (but maybe this already helps you. ;-) )
>
> Greetings,
> Peter
>
> ----- Original Message -----
> From: niklas@protocol7.com
> To: users@mina.apache.org
> Date: 03.08.2011 16:54:24
> Subject: Re: Vysper performance analysis
>
>
> > Very interesting, thanks for notifying us! I'm sure the german version
> > would be of interest as well, there are people around here who would
> > have no problem understanding it :-)
> >
> > Would you happen to have further details on the errors you got with
> > the high frequency test cases?
> >
> > /niklas
>
>
>
>
Re: Re: Vysper performance analysis
Posted by Peter Frohberg <em...@pre-dev.de>.
Hi Niklas,
I'm sorry but I really don't have any details that would help you. I did not try logging on debug level becaus the first 'test'-tests I ran Vysper with default (debug) logging I got the following error two times - before figuring out, that I should change the logging level:
java.net.SocketException: Broken pipe
at java.net.SocketOutputStream.socketWrite0(Native Method)
at java.net.SocketOutputStream.socketWrite(SocketOutputStream.java:92)
at java.net.SocketOutputStream.write(SocketOutputStream.java:136)
at com.sun.net.ssl.internal.ssl.OutputRecord.writeBuffer(OutputRecord.java:297)
at com.sun.net.ssl.internal.ssl.OutputRecord.write(OutputRecord.java:286)
at com.sun.net.ssl.internal.ssl.SSLSocketImpl.writeRecordInternal(SSLSocketImpl.java:743)
at com.sun.net.ssl.internal.ssl.SSLSocketImpl.writeRecord(SSLSocketImpl.java:731)
at com.sun.net.ssl.internal.ssl.AppOutputStream.write(AppOutputStream.java:59)
at sun.nio.cs.StreamEncoder.writeBytes(StreamEncoder.java:202)
at sun.nio.cs.StreamEncoder.implFlushBuffer(StreamEncoder.java:272)
at sun.nio.cs.StreamEncoder.implFlush(StreamEncoder.java:276)
at sun.nio.cs.StreamEncoder.flush(StreamEncoder.java:122)
at java.io.OutputStreamWriter.flush(OutputStreamWriter.java:212)
at java.io.BufferedWriter.flush(BufferedWriter.java:236)
at org.jivesoftware.smack.PacketWriter.writePackets(PacketWriter.java:193)
at org.jivesoftware.smack.PacketWriter.access$000(PacketWriter.java:40)
at org.jivesoftware.smack.PacketWriter$1.run(PacketWriter.java:76)
(but maybe this already helps you. ;-) )
Greetings,
Peter
----- Original Message -----
From: niklas@protocol7.com
To: users@mina.apache.org
Date: 03.08.2011 16:54:24
Subject: Re: Vysper performance analysis
> Very interesting, thanks for notifying us! I'm sure the german version
> would be of interest as well, there are people around here who would
> have no problem understanding it :-)
>
> Would you happen to have further details on the errors you got with
> the high frequency test cases?
>
> /niklas
Re: Vysper performance analysis
Posted by Niklas Gustavsson <ni...@protocol7.com>.
Very interesting, thanks for notifying us! I'm sure the german version
would be of interest as well, there are people around here who would
have no problem understanding it :-)
Would you happen to have further details on the errors you got with
the high frequency test cases?
/niklas
On Wed, Aug 3, 2011 at 4:41 PM, Peter Frohberg <em...@pre-dev.de> wrote:
> Hi all,
>
> I'm a computer scientist student at the German university TU Dresden.
>
> The past six months I wrote my minor thesis focusing on the scalibility of Vysper.
>
> Due the fact that the results are written in German, I tried to write a short summary in English, which I'd like to share and can be found here:
>
> http://www.pre-dev.de/vysper/SummaryAnalysisVysper2011.pdf
>
> If you have any questions redarding this summary, feel free to ask and I try my best to anwser them as detailed as possible.
>
> Greetings,
> Peter
>
>
>
>