I've been playing with ftpbench and betaftpd on and off, and sent betaftpd's author a few bugfixes for his daemon. Haven't had much time to enhance ftpbench...
Dan Kegel
dank@...
Feb 13, 2000 12:00 am
31
hi, You may need more than one computer acting as the ftp client to pull off 1000 connections to a server (But I wouldn't really know :-) ). You will also need...
Beau Kuiper
ekuiperba@...
Feb 13, 2000 12:05 am
32
Oh, you're right, no question. I'm getting around the disk problem by using a document tree of a single file small enough to fit in RAM, so there are no disk...
Dan Kegel
dank@...
Feb 13, 2000 12:24 am
33
... 'lack of disk threading'? Oh, do tell :) ... this only becomes a problem with lots of cpus and lots of interfaces.. you should be able to fill a gigabit...
Zach Brown
zab@...
Feb 13, 2000 2:51 pm
34
... Ops, I was wrong. 2.3.X improves SMP scalability of the disk subsystem. :-) Sorry for this mistake :-) ... Filling a gigabit pipe is not easy by any means,...
Beau Kuiper
ekuiperba@...
Feb 14, 2000 12:55 am
35
Hi, Under heavy load, strange things occur. System calls that have never failed may fail. Often these are not checked correctly and cause the server to crash. ...
Beau Kuiper
ekuiperba@...
Feb 13, 2000 12:56 am
36
Yeah, at one point I pointed out to the author that he really did need to check every system call's return value for errors. If there are still crashes going...
Dan Kegel
dank@...
Feb 14, 2000 1:33 am
37
... It also fixes the scalability in the network subsystem, _far_ more than in the disk. ... yes, tieing connection state to execution state is not the...
Zach Brown
zab@...
Feb 14, 2000 2:28 am
38
... But it makes writing programs easier to be honest :-) And generally safer. ... Well not exactly. The number of descriptors will always be a F(N) to the...
Beau Kuiper
ekuiperba@...
Feb 14, 2000 4:29 am
39
... this only is meaningful if the amount of data returned by your average poll is significantly related to the amount of data you fed in, which in practice is...
Zach Brown
zab@...
Feb 14, 2000 5:38 am
40
This doesn't add any functionality; it switches from a hand-written Makefile.in to one generated by automake, renames a file files and targets to conform to...
Dan Kegel
dank@...
Feb 21, 2000 6:39 am
41
Without any testing, I've released version 0.9 of dkftpbench. The only code change is a small one to handle_data_io, and is simply more correct error handling;...
Dan Kegel
dank@...
Mar 13, 2000 6:07 am
42
I did one final test before going to bed - I ran the client on a fast box, and the server on a slow one. It ran fine until it got to 126 clients: 126 users...
shane@...
Mar 13, 2000 6:42 am
43
... That's the best way to run it, actually, since dkftpbench is likely to be slower than a good ftp server. ... Don't be scared by the ugly log output -...
Dan Kegel
dank@...
Mar 13, 2000 12:45 pm
44
dkftpbench has helped me isolate a limit in the oftpd server. Right now oftpd creates two threads for each client connection: one to handle the control...
shane@...
Mar 14, 2000 5:19 am
45
... I suppose I could modify dklimits.c to probe to see how many threads can be created. (I wonder if that's dangerous?) ... The 2.4 kernel will raise that...
Dan Kegel
dank@...
Mar 14, 2000 6:57 am
46
Hi, I am going to be running some throughput test and I think dkftpbench will be very helpful. I will be running these test under Solaris, 2.5.1 and maybe 8....
Henry Baez
hbaez@...
Apr 20, 2000 6:19 pm
47
... It looks like my webserver didn't have the right MIME type for .tgz, so I changed the filename to .tar.gz. Let me know if you can download this link...
Dan Kegel
dank@...
Apr 20, 2000 7:24 pm
48
Hi Dan, The tar.gz downloaded OK. ./configure seem to work without any problem. The machine I am working on has GNU make version 3.71 however the make does...
Henry Baez
hbaez@...
Apr 21, 2000 2:07 pm
49
... If you email your Makefile and the output of make (e.g. 'make > log.txt', then email Makefile and log.txt) to me at dank@..., maybe I can ...
Dan Kegel
dank@...
Apr 21, 2000 3:49 pm
50
Hi all, I have been throwing a new anonymous ftp server design around in my head for the last couple of days. This design should easily be able to scale to...
Beau Kuiper
kuiperba@...
Jul 21, 2000 9:12 am
51
... Cool. I'd be glad to help, and would like to see a high-powered server as secure as oftpd. :) One concern: I'm not exactly sure what mincore does, but...
shane@...
Jul 24, 2000 10:43 am
52
... Thanks, When I have a detailed design ready, I will contact you for help coding/testing. It will: - Use chrooted environment - Possibly run non-root if I...
Beau Kuiper
kuiperba@...
Jul 24, 2000 1:09 pm
53
... Apache and Samba are not in the Insane Performance (tux, IIS, zeus, netscape directory services) category. They're in the actually useful real world ...
Zach Brown
zab@...
Jul 24, 2000 2:41 pm
54
... mincore is available in many flavors of Unix (including Linux since 2.3.51 or so); see http://www.kegel.com/c10k.html#nb It's a way to avoid blocking while...
Dan Kegel
dank@...
Jul 24, 2000 4:21 pm
55
... The LSB suggests using only standard pthreads threading, I think, since Linux will be moving towards full pthreads compliance as time goes on. I have to...
Dan Kegel
dank@...
Jul 24, 2000 4:22 pm
56
... I just checked ftp.cdrom.com, and it allows 5000 simultaneous connections. Right now there aren't any FTP servers on Linux that can handle this load, so...
shane@...
Jul 24, 2000 4:39 pm
57
... what it allows and what it can handle are two entirely different things ... machine can handle much more than that. ... multiple cpus and handling blocking...
Zach Brown
zab@...
Jul 24, 2000 5:45 pm
58
Summary: Let's see what kind of components the design requires and we can revisit the thread vs. process issue. Has the design been started (that is, has it...
shane@...
Jul 25, 2000 8:06 am
59
... Mutex results under linux on a celeron-333: N = 1, Linux => 1843325 operations/sec N = 2, Linux => 1843606 operations/sec N = 3, Linux => 1825140...