Opened 14 years ago

Last modified 6 years ago

#820 closed Bug report

FileZilla Server incredibly slow

Reported by: prezla Owned by: Alexander Schuch
Priority: normal Component: FileZilla Server
Keywords: Cc: prezla, Alexander Schuch, Tim Kosse
Component version: Operating system type:
Operating system version:

Description

FileZilla Server 0.9.5 is incredibly slow. I was
running it for and FTP server, but have since swapped
it out for WARFTPd 1.82.0.10 because of the following
performance numbers (notice that FileZilla is almost 3
times slower than WARFTPd, and Microsoft FTP is only
about 1 sec slower than WARFTPd):

FileZilla Server 0.9.5:


ftp> get catalog.pdf
200 Port command successful
150 Opening data channel for file transfer.
226 Transfer OK
ftp: 42509572 bytes received in 18.24Seconds
2330.95Kbytes/sec.

Microsoft FTP Server 5.1 (XP):


ftp> get catalog.pdf
200 PORT command successful.
150 Opening BINARY mode data connection for
catalog.pdf(42509572 bytes).
226 Transfer complete.
ftp: 42509572 bytes received in 8.14Seconds
5221.02Kbytes/sec.

WARFTPd 1.82.00-RC10:


ftp> get catalog.pdf
200 PORT command successful.
150 Opening BINARY mode data connection for catalog.pdf
(42509572 bytes).
226 Transfer complete. 42509572 bytes in 7.00 sec.
(5930.465 Kb/s)
ftp: 42509572 bytes received in 6.99Seconds
6080.61Kbytes/sec.

My platform is:
Intel Pentium III 933
512 MB RAM
Microsoft Windows XP SP2

Change History (3)

comment:1 Changed 14 years ago by Tim Kosse

Please have a look at the transfer buffer option in the
settings dialog. The default is 4096 which works best on
most systems. You may have to modify this value, especially
if you have modified any options of your TCP/IP stack or are
using the useless, so-called internet accellerators.

comment:2 Changed 14 years ago by prezla

Thanks for the response. Performance has improved
signicantly by increasing the transfer buffer to 16384.

Is there a reason why the default is 4096? I'm assuming
that the buffer is allocated for each connection. If this
is true, I would assume this is done to conserve memory?

comment:3 Changed 12 years ago by Tim Kosse

The default buffer sizes have since been increased.

Note: See TracTickets for help on using tickets.