You are viewing a plain text version of this content. The canonical link for it is here.
Posted to ftpserver-dev@incubator.apache.org by Niklas Gustavsson <ni...@protocol7.com> on 2007/01/06 15:23:52 UTC
Re: [jira] Commented: (FTPSERVER-48) Ability to change the size of
the buffer when writing data on disk
On 1/5/07, Nicolas FROMENT (JIRA) <ji...@apache.org> wrote:
> First of all, I was wrong the default size is 4096 and not 512. Some
time ago, I experiment performance issue with the ftpserver. It was 4
times slower than proftpd on the same computer. I look around the code.
As I was previously implicated in a project with a lot of IO call, I
knew that writing on disk with different buffer size (because of disk
cluster size and/or because bigger the buffer is, fewer are the IO) we
got different performance.
I would be interesting to test this with different sizes for the
buffer. And, I would guess that there are many other places where we
can optimize our code. I don't know if anyone has done any serious
performance work with FtpServer. If you like to do it, we would be
trilled :-)
> I don't test it with the ftpserver code (I can't manage building it
with maven... If someone can explain or point to a wiki page :-) ). With
the last binaries I download, it is still 4 times slower than proftpd.
The binaries has not been updated for a few years so there no change there.
This, rather brief, page in the wiki describes what you should need to
build FtpServer:
http://cwiki.apache.org/FTPSERVER/building.html
If you get stuck anywhere doing it, feel free to ask here and I'll help
you out!
/niklas
Re: [jira] Commented: (FTPSERVER-48) Ability to change the size of
the buffer when writing data on disk
Posted by Nicolas FROMENT <nf...@histor.fr>.
Great ! One last thing, the build failed because maven doesn't find a
test framework to test the admin gui component. I don't if unit test are
needed on this component but adding junit dependencies to the admin gui
pom maje the build successful.
Niklas Gustavsson a écrit :
> Nicolas FROMENT wrote:
>> Ok. I run maven from the command line instead of in eclipse. And it's
>> better but no perfect.
>> I got two errors.
>> 1)In the class: PassivePorts.java, you use Integer.valueOf(int) on
>> line 110. I guess this is a mistake and you want to write new
>> Integer(int)
>
> Sorry, again my bad. Integer.valueOf(int) requires Java 1.5. Fixed in
> trunk now. Again, we really need that CI server :-)
>
>> 2)One I resolve the first issue the core compiles, but the gui still
>> have errors. I guess that is related to the last post of Mark
>
> Yes, this is probably due to the same reason as Mark pointed out
> (which should now have been fixed).
>
> /niklas
>
>
--
******************************************************************
Nicolas Froment nfroment@histor.fr
Software Architect
Hi-Stor Technologies www.histor.fr
23 bd Victor Hugo Std: +33 (0) 562 12 14 40 (Voice)
Bâtiment Socrate Fax: +33 (0) 562 12 14 49
31770 Colomiers
FRANCE
****************************************************************
Re: [jira] Commented: (FTPSERVER-48) Ability to change the size of
the buffer when writing data on disk
Posted by Niklas Gustavsson <ni...@protocol7.com>.
Nicolas FROMENT wrote:
> Ok. I run maven from the command line instead of in eclipse. And it's
> better but no perfect.
> I got two errors.
> 1)In the class: PassivePorts.java, you use Integer.valueOf(int) on line
> 110. I guess this is a mistake and you want to write new Integer(int)
Sorry, again my bad. Integer.valueOf(int) requires Java 1.5. Fixed in
trunk now. Again, we really need that CI server :-)
> 2)One I resolve the first issue the core compiles, but the gui still
> have errors. I guess that is related to the last post of Mark
Yes, this is probably due to the same reason as Mark pointed out (which
should now have been fixed).
/niklas
Re: [jira] Commented: (FTPSERVER-48) Ability to change the size of
the buffer when writing data on disk
Posted by Nicolas FROMENT <nf...@histor.fr>.
Ok. I run maven from the command line instead of in eclipse. And it's
better but no perfect.
I got two errors.
1)In the class: PassivePorts.java, you use Integer.valueOf(int) on line
110. I guess this is a mistake and you want to write new Integer(int)
2)One I resolve the first issue the core compiles, but the gui still
have errors. I guess that is related to the last post of Mark
Niklas Gustavsson a écrit :
> On 1/5/07, Nicolas FROMENT (JIRA) <ji...@apache.org> wrote:
> > First of all, I was wrong the default size is 4096 and not 512. Some
> time ago, I experiment performance issue with the ftpserver. It was 4
> times slower than proftpd on the same computer. I look around the
> code. As I was previously implicated in a project with a lot of IO
> call, I knew that writing on disk with different buffer size (because
> of disk cluster size and/or because bigger the buffer is, fewer are
> the IO) we got different performance.
>
> I would be interesting to test this with different sizes for the
> buffer. And, I would guess that there are many other places where we
> can optimize our code. I don't know if anyone has done any serious
> performance work with FtpServer. If you like to do it, we would be
> trilled :-)
>
> > I don't test it with the ftpserver code (I can't manage building it
> with maven... If someone can explain or point to a wiki page :-) ).
> With the last binaries I download, it is still 4 times slower than
> proftpd.
>
> The binaries has not been updated for a few years so there no change
> there.
>
> This, rather brief, page in the wiki describes what you should need to
> build FtpServer:
> http://cwiki.apache.org/FTPSERVER/building.html
>
> If you get stuck anywhere doing it, feel free to ask here and I'll
> help you out!
>
> /niklas
>
>
--
******************************************************************
Nicolas Froment nfroment@histor.fr
Software Architect
Hi-Stor Technologies www.histor.fr
23 bd Victor Hugo Std: +33 (0) 562 12 14 40 (Voice)
Bâtiment Socrate Fax: +33 (0) 562 12 14 49
31770 Colomiers
FRANCE
****************************************************************