java performance w2k vs NT4

[Bastien Raich]

Hello,

I've been developping a Java application using sun's 1.4.1 on my NT4
box.

It's a financial application that retrieves market data in real time,
displays it in JTables, do some computations, and from which you can
also open small dialogs to send orders to the market... client /
server based (but not Java connection, it's a simple socket
connection)

my clients have been running NT4 for some time, and we recently
decided to migrate to windows 2000. after all I had read about java
portability and performances of w2k vs NT, I just quickly checked the
app was starting up, and started the migration.

however after we finished the first PC and ran it the next day in
production, we realized everything was REALLY slow. I mean MUCH slower
than on the other PCs still running NT4. I could not tell for sure (my
users are remote), but CPU usage seems to be more intensive on the w2k
box.

another interesting thing was that things are REALLY slow with jre
1.4.2 and 1.4.1, but "quite ok, although slower" with jre 1.4.0 .

I am really puzzled... the machines used are of pretty similar specs.

so do you know of any perf. issue between those two OSes ?
any specific code I should pay more attention to ?
any known problem with w2k ? I searched the NG & knowledge bases with
no success...

thanks to share your knowledge !

--Bastien

[Bastien Raich]

more infos:
I recreated the problem by using another application of my package: it
retrieve info from a server, and display it line by line in JTables. I
c

I compared the time it takes to retrieve all the info, it's < 20
seconds on my NT machine, while it's > 60 seconds on the w2k machine.
the NT machine is less powerful and runs more apps than the w2k.

It looks like being a problem at the display level, as when I minimize
my window, the whole info is updated in around 20 secs on the w2k
machine (same as NT).

has anyone seen this already ?
I haven't tested with linux, that might be an idea...

sr****@clj.co.jp (Bastien Raich) wrote in message news:<ae**************************@posting.google. com>...

Hello,

I've been developping a Java application using sun's 1.4.1 on my NT4
box.

It's a financial application that retrieves market data in real time,
displays it in JTables, do some computations, and from which you can
also open small dialogs to send orders to the market... client /
server based (but not Java connection, it's a simple socket
connection)

my clients have been running NT4 for some time, and we recently
decided to migrate to windows 2000. after all I had read about java
portability and performances of w2k vs NT, I just quickly checked the
app was starting up, and started the migration.

however after we finished the first PC and ran it the next day in
production, we realized everything was REALLY slow. I mean MUCH slower
than on the other PCs still running NT4. I could not tell for sure (my
users are remote), but CPU usage seems to be more intensive on the w2k
box.

another interesting thing was that things are REALLY slow with jre
1.4.2 and 1.4.1, but "quite ok, although slower" with jre 1.4.0 .

I am really puzzled... the machines used are of pretty similar specs.

so do you know of any perf. issue between those two OSes ?
any specific code I should pay more attention to ?
any known problem with w2k ? I searched the NG & knowledge bases with
no success...

thanks to share your knowledge !

--Bastien

[nos]

Have you tried using "windows task manager"?
You can look at the performance window which
gives you cpu usage and page file usage charts, and various
other things like physical/kernel memory, handles,
threads and processes. The process window gives you
cpu and memory usage for each process. 2000 may
have a networking window.
"Bastien Raich" <sr****@clj.co.jp> wrote in message
news:ae**************************@posting.google.c om...

more infos:
I recreated the problem by using another application of my package: it
retrieve info from a server, and display it line by line in JTables. I
c

I compared the time it takes to retrieve all the info, it's < 20
seconds on my NT machine, while it's > 60 seconds on the w2k machine.
the NT machine is less powerful and runs more apps than the w2k.

It looks like being a problem at the display level, as when I minimize
my window, the whole info is updated in around 20 secs on the w2k
machine (same as NT).

has anyone seen this already ?
I haven't tested with linux, that might be an idea...

sr****@clj.co.jp (Bastien Raich) wrote in message

news:<ae**************************@posting.google. com>...

Hello,

I've been developping a Java application using sun's 1.4.1 on my NT4
box.

It's a financial application that retrieves market data in real time,
displays it in JTables, do some computations, and from which you can
also open small dialogs to send orders to the market... client /
server based (but not Java connection, it's a simple socket
connection)

my clients have been running NT4 for some time, and we recently
decided to migrate to windows 2000. after all I had read about java
portability and performances of w2k vs NT, I just quickly checked the
app was starting up, and started the migration.

however after we finished the first PC and ran it the next day in
production, we realized everything was REALLY slow. I mean MUCH slower
than on the other PCs still running NT4. I could not tell for sure (my
users are remote), but CPU usage seems to be more intensive on the w2k
box.

another interesting thing was that things are REALLY slow with jre
1.4.2 and 1.4.1, but "quite ok, although slower" with jre 1.4.0 .

I am really puzzled... the machines used are of pretty similar specs.

so do you know of any perf. issue between those two OSes ?
any specific code I should pay more attention to ?
any known problem with w2k ? I searched the NG & knowledge bases with
no success...

thanks to share your knowledge !

--Bastien

[Bastien Raich]

of course I had a look at the task manager, CPU usage is higher with windows 2000.

I upgraded graphic card drivers, but it didn't change anything...

any idea ?

--Bastien

"nos" <no*@nospam.com> wrote in message news:<C2avb.256408$HS4.2307827@attbi_s01>...

Have you tried using "windows task manager"?
You can look at the performance window which
gives you cpu usage and page file usage charts, and various
other things like physical/kernel memory, handles,
threads and processes. The process window gives you
cpu and memory usage for each process. 2000 may
have a networking window.
"Bastien Raich" <sr****@clj.co.jp> wrote in message
news:ae**************************@posting.google.c om...

more infos:
I recreated the problem by using another application of my package: it
retrieve info from a server, and display it line by line in JTables. I
c

I compared the time it takes to retrieve all the info, it's < 20
seconds on my NT machine, while it's > 60 seconds on the w2k machine.
the NT machine is less powerful and runs more apps than the w2k.

It looks like being a problem at the display level, as when I minimize
my window, the whole info is updated in around 20 secs on the w2k
machine (same as NT).

has anyone seen this already ?
I haven't tested with linux, that might be an idea...

sr****@clj.co.jp (Bastien Raich) wrote in message

news:<ae**************************@posting.google. com>...

Hello,

I've been developping a Java application using sun's 1.4.1 on my NT4
box.

It's a financial application that retrieves market data in real time,
displays it in JTables, do some computations, and from which you can
also open small dialogs to send orders to the market... client /
server based (but not Java connection, it's a simple socket
connection)

my clients have been running NT4 for some time, and we recently
decided to migrate to windows 2000. after all I had read about java
portability and performances of w2k vs NT, I just quickly checked the
app was starting up, and started the migration.

however after we finished the first PC and ran it the next day in
production, we realized everything was REALLY slow. I mean MUCH slower
than on the other PCs still running NT4. I could not tell for sure (my
users are remote), but CPU usage seems to be more intensive on the w2k
box.

another interesting thing was that things are REALLY slow with jre
1.4.2 and 1.4.1, but "quite ok, although slower" with jre 1.4.0 .

I am really puzzled... the machines used are of pretty similar specs.

so do you know of any perf. issue between those two OSes ?
any specific code I should pay more attention to ?
any known problem with w2k ? I searched the NG & knowledge bases with
no success...

thanks to share your knowledge !

--Bastien

[Larry]

Try installing your app on another PC with w2k and see if you get the same
results. I wouldn't base your findings on one installation.....

Larry

"Bastien Raich" <sr****@clj.co.jp> wrote in message
news:ae**************************@posting.google.c om...

Hello,

I've been developping a Java application using sun's 1.4.1 on my NT4
box.

It's a financial application that retrieves market data in real time,
displays it in JTables, do some computations, and from which you can
also open small dialogs to send orders to the market... client /
server based (but not Java connection, it's a simple socket
connection)

my clients have been running NT4 for some time, and we recently
decided to migrate to windows 2000. after all I had read about java
portability and performances of w2k vs NT, I just quickly checked the
app was starting up, and started the migration.

however after we finished the first PC and ran it the next day in
production, we realized everything was REALLY slow. I mean MUCH slower
than on the other PCs still running NT4. I could not tell for sure (my
users are remote), but CPU usage seems to be more intensive on the w2k
box.

another interesting thing was that things are REALLY slow with jre
1.4.2 and 1.4.1, but "quite ok, although slower" with jre 1.4.0 .

I am really puzzled... the machines used are of pretty similar specs.

so do you know of any perf. issue between those two OSes ?
any specific code I should pay more attention to ?
any known problem with w2k ? I searched the NG & knowledge bases with
no success...

thanks to share your knowledge !

--Bastien