You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@mina.apache.org by Steve Johns <st...@gmail.com> on 2008/04/09 17:00:27 UTC
mina2.0.0-M1 vs mina1.1.6
I used both mina1.1.6 and mina2.0.0-M1(replaced mina1.1.6 and didn't change
any other codes) to make two servers which are both connected to same data
source. Two same clients connected to these two servers. I observed the
client connected mina2.0.0-M1 receiving much slower(maybe missing) than
mina1.1.6 under heavy loading. (8000messages/sec). WhenI stopped playing my
data source, client A received less messages than client B did. All programs
were on the same machine. I wonder if any critical algorithmn was changed in
mina2.0.0-M1? I saw one thread mentioning about M2 write slow issue. Any
idea?
RE: mina2.0.0-M1 vs mina1.1.6
Posted by "Craig L. Ching" <cc...@mqsoftware.com>.
I don't know if it's related or not, but a bug was fixed recently where
byte buffer expansion was much slower than it should have been. If
you're using expandable byte buffers, you might want to check out the
svn trunk and test with that.
Cheers,
Craig
> -----Original Message-----
> From: Steve Johns [mailto:steven.mark.johns@gmail.com]
> Sent: Wednesday, April 09, 2008 10:00 AM
> To: mina
> Subject: mina2.0.0-M1 vs mina1.1.6
>
> I used both mina1.1.6 and mina2.0.0-M1(replaced mina1.1.6 and didn't
> change
> any other codes) to make two servers which are both connected to same
data
> source. Two same clients connected to these two servers. I observed
the
> client connected mina2.0.0-M1 receiving much slower(maybe missing)
than
> mina1.1.6 under heavy loading. (8000messages/sec). WhenI stopped
playing
> my
> data source, client A received less messages than client B did. All
> programs
> were on the same machine. I wonder if any critical algorithmn was
changed
> in
> mina2.0.0-M1? I saw one thread mentioning about M2 write slow issue.
Any
> idea?
Re: mina2.0.0-M1 vs mina1.1.6
Posted by Steve Johns <st...@gmail.com>.
Hi Trustin
Is it possible to do a testing on Mina2.0.0-M1 under heavy loading? The
throughput value (messages/sec) somehow can't tell either messages deliver
are dealy or lost. I am sorry that I can't post my code for reproducing the
problems since the data source server data is too huge to publish to
internet and also not allowed. Thanks again.
2008/4/10 Steve Johns <st...@gmail.com>:
> OK. I did something.
> 1) I used two clients connecting to server with 1.1.6. Two clients
> received the same amount of messages eventually after I stopped the data
> source server.
> 2) Two clients to server with 2.0.0.-M1. two clients received messages
> numbers are NOT same. (packet lost? if slow, two clients eventually should
> receive the same amounts of messages when I stopped the data source server,
> right?)
>
> 2008/4/10 Steve Johns <st...@gmail.com>:
>
>
> >
> > On Thu, Apr 10, 2008 at 12:16 AM, "이희승 (Trustin Lee) <tr...@gmail.com>
> > wrote:
> >
> > > It sounds like MINA 2 performs better when there's no data source
> > > server
> > > access. Did I get it right? :)
> >
> > No. Data source server is just a server which could generate huge
> > amounts of messages. Actually I don't know how to simulate creating a big
> > throughput messages and i simply connect to other's server. ^^ However. My
> > way can't tell if it is IoConnector problem or IoAcceptor problem without
> > further testing.
> >
> > >
> > > Then.. that's very weird. What do you use to communicate with the
> > > data
> > > source server? Do you use MINA there too (i.e. IoConnector in
> > > IoHandler)?
> >
> > Yes
> >
> > >
> > >
> > > Steve Johns wrote:
> > > > (FASTER) 1. Client A -> Server with mina1.1.6 -> data
> > > > source server C (8000messages/sec)
> > > > (much SLOWER) 2. Client B -> Server with mina2.0.0 -> data source
> > > server C
> > > > I may try more clients and tell you the result.
> > > > On Thu, Apr 10, 2008 at 12:04 AM, "이희승 (Trustin Lee)
> > > > <trustin@gmail.com <ma...@gmail.com>> wrote:
> > > >
> > > > Which client was connected MINA 2, Client A or B?
> > > >
> > > > There was huge refactoring in MINA 2, so there can be
> > > differences in
> > > > performance characteristics. However, I don't think I have
> > > changed any
> > > > critical algorithm.
> > > >
> > > > What would happen if you increase the number of simultaneous
> > > clients?
> > > >
> > > > Steve Johns wrote:
> > > > > I used both mina1.1.6 and mina2.0.0-M1(replaced mina1.1.6 and
> > > > didn't change
> > > > > any other codes) to make two servers which are both connected
> > > to
> > > > same data
> > > > > source. Two same clients connected to these two servers. I
> > > > observed the
> > > > > client connected mina2.0.0-M1 receiving much slower(maybe
> > > missing)
> > > > than
> > > > > mina1.1.6 under heavy loading. (8000messages/sec). WhenI
> > > stopped
> > > > playing my
> > > > > data source, client A received less messages than client B
> > > did.
> > > > All programs
> > > > > were on the same machine. I wonder if any critical algorithmn
> > > was
> > > > changed in
> > > > > mina2.0.0-M1? I saw one thread mentioning about M2 write slow
> > > > issue. Any
> > > > > idea?
> > > > >
> > > >
> > > > --
> > > > Trustin Lee - Principal Software Engineer, JBoss, Red Hat
> > > > --
> > > > what we call human nature is actually human habit
> > > > --
> > > > http://gleamynode.net/
> > > >
> > > >
> > >
> > > --
> > > Trustin Lee - Principal Software Engineer, JBoss, Red Hat
> > > --
> > > what we call human nature is actually human habit
> > > --
> > > http://gleamynode.net/
> > >
> > >
> >
>
Re: mina2.0.0-M1 vs mina1.1.6
Posted by Steve Johns <st...@gmail.com>.
OK. I did something.
1) I used two clients connecting to server with 1.1.6. Two clients received
the same amount of messages eventually after I stopped the data source
server.
2) Two clients to server with 2.0.0.-M1. two clients received messages
numbers are NOT same. (packet lost? if slow, two clients eventually should
receive the same amounts of messages when I stopped the data source server,
right?)
2008/4/10 Steve Johns <st...@gmail.com>:
>
>
> On Thu, Apr 10, 2008 at 12:16 AM, "이희승 (Trustin Lee) <tr...@gmail.com>
> wrote:
>
> > It sounds like MINA 2 performs better when there's no data source server
> > access. Did I get it right? :)
>
> No. Data source server is just a server which could generate huge amounts
> of messages. Actually I don't know how to simulate creating a big throughput
> messages and i simply connect to other's server. ^^ However. My way can't
> tell if it is IoConnector problem or IoAcceptor problem without further
> testing.
>
> >
> > Then.. that's very weird. What do you use to communicate with the data
> > source server? Do you use MINA there too (i.e. IoConnector in
> > IoHandler)?
>
> Yes
>
> >
> >
> > Steve Johns wrote:
> > > (FASTER) 1. Client A -> Server with mina1.1.6 -> data
> > > source server C (8000messages/sec)
> > > (much SLOWER) 2. Client B -> Server with mina2.0.0 -> data source
> > server C
> > > I may try more clients and tell you the result.
> > > On Thu, Apr 10, 2008 at 12:04 AM, "이희승 (Trustin Lee)
> > > <trustin@gmail.com <ma...@gmail.com>> wrote:
> > >
> > > Which client was connected MINA 2, Client A or B?
> > >
> > > There was huge refactoring in MINA 2, so there can be differences
> > in
> > > performance characteristics. However, I don't think I have
> > changed any
> > > critical algorithm.
> > >
> > > What would happen if you increase the number of simultaneous
> > clients?
> > >
> > > Steve Johns wrote:
> > > > I used both mina1.1.6 and mina2.0.0-M1(replaced mina1.1.6 and
> > > didn't change
> > > > any other codes) to make two servers which are both connected to
> > > same data
> > > > source. Two same clients connected to these two servers. I
> > > observed the
> > > > client connected mina2.0.0-M1 receiving much slower(maybe
> > missing)
> > > than
> > > > mina1.1.6 under heavy loading. (8000messages/sec). WhenI stopped
> > > playing my
> > > > data source, client A received less messages than client B did.
> > > All programs
> > > > were on the same machine. I wonder if any critical algorithmn
> > was
> > > changed in
> > > > mina2.0.0-M1? I saw one thread mentioning about M2 write slow
> > > issue. Any
> > > > idea?
> > > >
> > >
> > > --
> > > Trustin Lee - Principal Software Engineer, JBoss, Red Hat
> > > --
> > > what we call human nature is actually human habit
> > > --
> > > http://gleamynode.net/
> > >
> > >
> >
> > --
> > Trustin Lee - Principal Software Engineer, JBoss, Red Hat
> > --
> > what we call human nature is actually human habit
> > --
> > http://gleamynode.net/
> >
> >
>
Re: mina2.0.0-M1 vs mina1.1.6
Posted by Steve Johns <st...@gmail.com>.
On Thu, Apr 10, 2008 at 12:16 AM, "이희승 (Trustin Lee) <tr...@gmail.com>
wrote:
> It sounds like MINA 2 performs better when there's no data source server
> access. Did I get it right? :)
No. Data source server is just a server which could generate huge amounts of
messages. Actually I don't know how to simulate creating a big throughput
messages and i simply connect to other's server. ^^ However. My way can't
tell if it is IoConnector problem or IoAcceptor problem without further
testing.
>
> Then.. that's very weird. What do you use to communicate with the data
> source server? Do you use MINA there too (i.e. IoConnector in IoHandler)?
Yes
>
>
> Steve Johns wrote:
> > (FASTER) 1. Client A -> Server with mina1.1.6 -> data
> > source server C (8000messages/sec)
> > (much SLOWER) 2. Client B -> Server with mina2.0.0 -> data source
> server C
> > I may try more clients and tell you the result.
> > On Thu, Apr 10, 2008 at 12:04 AM, "이희승 (Trustin Lee)
> > <trustin@gmail.com <ma...@gmail.com>> wrote:
> >
> > Which client was connected MINA 2, Client A or B?
> >
> > There was huge refactoring in MINA 2, so there can be differences in
> > performance characteristics. However, I don't think I have changed
> any
> > critical algorithm.
> >
> > What would happen if you increase the number of simultaneous
> clients?
> >
> > Steve Johns wrote:
> > > I used both mina1.1.6 and mina2.0.0-M1(replaced mina1.1.6 and
> > didn't change
> > > any other codes) to make two servers which are both connected to
> > same data
> > > source. Two same clients connected to these two servers. I
> > observed the
> > > client connected mina2.0.0-M1 receiving much slower(maybe missing)
> > than
> > > mina1.1.6 under heavy loading. (8000messages/sec). WhenI stopped
> > playing my
> > > data source, client A received less messages than client B did.
> > All programs
> > > were on the same machine. I wonder if any critical algorithmn was
> > changed in
> > > mina2.0.0-M1? I saw one thread mentioning about M2 write slow
> > issue. Any
> > > idea?
> > >
> >
> > --
> > Trustin Lee - Principal Software Engineer, JBoss, Red Hat
> > --
> > what we call human nature is actually human habit
> > --
> > http://gleamynode.net/
> >
> >
>
> --
> Trustin Lee - Principal Software Engineer, JBoss, Red Hat
> --
> what we call human nature is actually human habit
> --
> http://gleamynode.net/
>
>
Re: mina2.0.0-M1 vs mina1.1.6
Posted by 이희승 (Trustin Lee) <t...@gmail.com>.
It sounds like MINA 2 performs better when there's no data source server
access. Did I get it right? :)
Then.. that's very weird. What do you use to communicate with the data
source server? Do you use MINA there too (i.e. IoConnector in IoHandler)?
Steve Johns wrote:
> (FASTER) 1. Client A -> Server with mina1.1.6 -> data
> source server C (8000messages/sec)
> (much SLOWER) 2. Client B -> Server with mina2.0.0 -> data source server C
> I may try more clients and tell you the result.
> On Thu, Apr 10, 2008 at 12:04 AM, "이희승 (Trustin Lee)
> <trustin@gmail.com <ma...@gmail.com>> wrote:
>
> Which client was connected MINA 2, Client A or B?
>
> There was huge refactoring in MINA 2, so there can be differences in
> performance characteristics. However, I don't think I have changed any
> critical algorithm.
>
> What would happen if you increase the number of simultaneous clients?
>
> Steve Johns wrote:
> > I used both mina1.1.6 and mina2.0.0-M1(replaced mina1.1.6 and
> didn't change
> > any other codes) to make two servers which are both connected to
> same data
> > source. Two same clients connected to these two servers. I
> observed the
> > client connected mina2.0.0-M1 receiving much slower(maybe missing)
> than
> > mina1.1.6 under heavy loading. (8000messages/sec). WhenI stopped
> playing my
> > data source, client A received less messages than client B did.
> All programs
> > were on the same machine. I wonder if any critical algorithmn was
> changed in
> > mina2.0.0-M1? I saw one thread mentioning about M2 write slow
> issue. Any
> > idea?
> >
>
> --
> Trustin Lee - Principal Software Engineer, JBoss, Red Hat
> --
> what we call human nature is actually human habit
> --
> http://gleamynode.net/
>
>
--
Trustin Lee - Principal Software Engineer, JBoss, Red Hat
--
what we call human nature is actually human habit
--
http://gleamynode.net/
Re: mina2.0.0-M1 vs mina1.1.6
Posted by Steve Johns <st...@gmail.com>.
(FASTER) 1. Client A -> Server with mina1.1.6 -> data
source server C (8000messages/sec)
(much SLOWER) 2. Client B -> Server with mina2.0.0 -> data source server C
I may try more clients and tell you the result.
On Thu, Apr 10, 2008 at 12:04 AM, "이희승 (Trustin Lee) <tr...@gmail.com>
wrote:
> Which client was connected MINA 2, Client A or B?
>
> There was huge refactoring in MINA 2, so there can be differences in
> performance characteristics. However, I don't think I have changed any
> critical algorithm.
>
> What would happen if you increase the number of simultaneous clients?
>
> Steve Johns wrote:
> > I used both mina1.1.6 and mina2.0.0-M1(replaced mina1.1.6 and didn't
> change
> > any other codes) to make two servers which are both connected to same
> data
> > source. Two same clients connected to these two servers. I observed the
> > client connected mina2.0.0-M1 receiving much slower(maybe missing) than
> > mina1.1.6 under heavy loading. (8000messages/sec). WhenI stopped playing
> my
> > data source, client A received less messages than client B did. All
> programs
> > were on the same machine. I wonder if any critical algorithmn was
> changed in
> > mina2.0.0-M1? I saw one thread mentioning about M2 write slow issue. Any
> > idea?
> >
>
> --
> Trustin Lee - Principal Software Engineer, JBoss, Red Hat
> --
> what we call human nature is actually human habit
> --
> http://gleamynode.net/
>
>
Re: mina2.0.0-M1 vs mina1.1.6
Posted by 이희승 (Trustin Lee) <t...@gmail.com>.
Which client was connected MINA 2, Client A or B?
There was huge refactoring in MINA 2, so there can be differences in
performance characteristics. However, I don't think I have changed any
critical algorithm.
What would happen if you increase the number of simultaneous clients?
Steve Johns wrote:
> I used both mina1.1.6 and mina2.0.0-M1(replaced mina1.1.6 and didn't change
> any other codes) to make two servers which are both connected to same data
> source. Two same clients connected to these two servers. I observed the
> client connected mina2.0.0-M1 receiving much slower(maybe missing) than
> mina1.1.6 under heavy loading. (8000messages/sec). WhenI stopped playing my
> data source, client A received less messages than client B did. All programs
> were on the same machine. I wonder if any critical algorithmn was changed in
> mina2.0.0-M1? I saw one thread mentioning about M2 write slow issue. Any
> idea?
>
--
Trustin Lee - Principal Software Engineer, JBoss, Red Hat
--
what we call human nature is actually human habit
--
http://gleamynode.net/