Skip to contentSkip to conversation

What is 2.4 Linux networking performance like compared to BSD?

DISCUSSION Hans Reiser 14 participants 22 messages Mar 1, 2001 – Mar 5, 2001

lore
Share & tools
lore
#0 lore raw

I have a client that wants to implement a webcache, but is very leery of implementing it on Linux rather than BSD.

They know that iMimic's polymix performance on Linux 2.2.* is half what it is on BSD. Has the Linux 2.4 networking code caught up to BSD?

Can I tell them not to worry about the Linux networking code strangling their webcache product's performance, or not?

Hans

#1 lore raw

hans,

we've found that the TCP and UDP performance on 2.4 is *dramatically* better than 2.2. with the acenic gig-e driver on PIII-933 UP (66MHz x 64bits PCI) we are getting 993 Mb/s with 2.4.0 with jumbo frames (about 850 Mb/s with standard ethernet frames). the best number we got with 2.2 was about 650 with jumbos and 550 with standard.

i'd recommend it's networking performance to anyone.

todd underwood
[email protected]

Quoted 21 lines · Hans Reiser

Date: Thu, 01 Mar 2001 02:26:20 +0300
From: Hans Reiser <[email protected]>
To: "[email protected]" <[email protected]>
Subject: What is 2.4 Linux networking performance like compared to BSD?

I have a client that wants to implement a webcache, but is very leery of implementing it on Linux rather than BSD.

They know that iMimic's polymix performance on Linux 2.2.* is half what it is on BSD. Has the Linux 2.4 networking code caught up to BSD?

Can I tell them not to worry about the Linux networking code strangling their webcache product's performance, or not?

Hans
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/

#15 lore raw
In reply to Hans Reiser
Quoted 2 lines

They know that iMimic's polymix performance on Linux 2.2.* is half what it is on BSD. Has the Linux 2.4 networking code caught up to BSD?

For the general case its been the same sort of speed as BSD sometimes faster sometimes not since before 2.2. So the problem they see would need more analysis to understand the real cause

#2 lore raw
In reply to Todd
Quoted 9 lines · Todd

hans,

we've found that the TCP and UDP performance on 2.4 is *dramatically* better than 2.2. with the acenic gig-e driver on PIII-933 UP (66MHz x 64bits PCI) we are getting 993 Mb/s with 2.4.0 with jumbo frames (about 850 Mb/s with standard ethernet frames). the best number we got with 2.2 was about 650 with jumbos and 550 with standard.

Quoted 28 lines

i'd recommend it's networking performance to anyone.

todd underwood
[email protected]

Quoted 21 lines · Hans Reiser

Date: Thu, 01 Mar 2001 02:26:20 +0300
From: Hans Reiser <[email protected]>
To: "[email protected]" <[email protected]>
Subject: What is 2.4 Linux networking performance like compared to BSD?

I have a client that wants to implement a webcache, but is very leery of implementing it on Linux rather than BSD.

They know that iMimic's polymix performance on Linux 2.2.* is half what it is on BSD. Has the Linux 2.4 networking code caught up to BSD?

Can I tell them not to worry about the Linux networking code strangling their webcache product's performance, or not?

Hans
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/

The problem is that I really need BSD vs. Linux experiences, not Linux 2.4 vs. 2.2 experiences, because the webcache industry tends to strongly disparage Linux networking code, so much better isn't necessarily good enough.

Hans

#5 lore raw
In reply to Todd
Quoted 5 lines · Hans Reiser

The problem is that I really need BSD vs. Linux experiences, not Linux 2.4 vs. 2.2 experiences, because the webcache industry tends to strongly disparage Linux networking code, so much better isn't necessarily good enough.

Hans

Check with the www.swelltech.com people, they should have the info you need.

http://www.swelltech.com/pengies/joe/squidtuneup/t1.html

The above link contains some decent squid performance hints for 2.2+Squid.

-Nathan Dabney

#3 lore raw
In reply to Hans Reiser
Quoted 8 lines · Hans Reiser

Date: Thu, 01 Mar 2001 19:03:31 +0300

Quoted 4 lines · Todd

hans,

we've found that the TCP and UDP performance on 2.4 is *dramatically* better than 2.2.

[..]

Quoted 9 lines
Quoted 8 lines

i'd recommend it's networking performance to anyone.

Quoted 2 lines · Hans Reiser

They know that iMimic's polymix performance on Linux 2.2.* is half what it is on BSD. Has the Linux 2.4 networking code caught up to BSD?

Quoted 8 lines
Quoted 2 lines
Quoted 2 lines

Can I tell them not to worry about the Linux networking code strangling their webcache product's performance, or not?

The problem is that I really need BSD vs. Linux experiences, not Linux 2.4 vs. 2.2 experiences, because the webcache industry tends to strongly disparage Linux networking code, so much better isn't necessarily good enough.

It isn't just the webcache industry, heh. I have not yet played with 2.4, let alone under what I consider stress; but from experience with 2.2 and eairlier I could see why one would take fbsd over linux. Between mysterious messages popping up on the console (be it they are related to NIC drivers or not), and other oddities as ram and fd's fill up, fbsd could be considered by some to be better suited.

On the topic of perfromance, I see Todd and a few others post some numbers, but has anyone kept track of them through kernel versions and
drivers? It would be interesting to see something like lmbench run on
each, and their results recorded.

I'm tempted to run various tests before and after I upgrade from 2.2.x to 2.4.x, just to see the difference ....

</crazy>

#16 lore raw
In reply to Hans Reiser
Quoted 8 lines · Hans Reiser

I have a client that wants to implement a webcache, but is very leery of implementing it on Linux rather than BSD.

They know that iMimic's polymix performance on Linux 2.2.* is half what it is on BSD. Has the Linux 2.4 networking code caught up to BSD?

Can I tell them not to worry about the Linux networking code strangling their webcache product's performance, or not?

Hi Hans,
I dont have an answer for you, but it would be nice to know the answer.
Would it be difficult to measure this? It should not be difficult to make a machine dual boot Linux and BSD, and then we can measure the differences. If there is a significant performance difference either way then we can try and investigate it to see why.

Thanks,

Jim

#6 lore raw
In reply to Todd
Quoted 8 lines · Nathan Dabney
Quoted 5 lines · Hans Reiser

The problem is that I really need BSD vs. Linux experiences, not Linux 2.4 vs. 2.2 experiences, because the webcache industry tends to strongly disparage Linux networking code, so much better isn't necessarily good enough.

Hans

Quoted 7 lines

Check with the www.swelltech.com people, they should have the info you need.

http://www.swelltech.com/pengies/joe/squidtuneup/t1.html

The above link contains some decent squid performance hints for 2.2+Squid.

-Nathan Dabney

It does not say anything about BSD vs. Linux 2.4 networking code.

If I can't get information about BSD v. Linux 2.4 networking code, then reiserfs has to get ported to BSD which will be both nice and a pain to do.

Hans

#17 lore raw
In reply to Hans Reiser
Quoted 18 lines · James Lewis Nance
Quoted 8 lines · Hans Reiser

I have a client that wants to implement a webcache, but is very leery of implementing it on Linux rather than BSD.

They know that iMimic's polymix performance on Linux 2.2.* is half what it is on BSD. Has the Linux 2.4 networking code caught up to BSD?

Can I tell them not to worry about the Linux networking code strangling their webcache product's performance, or not?

Hi Hans,
I dont have an answer for you, but it would be nice to know the answer.
Would it be difficult to measure this? It should not be difficult to make a machine dual boot Linux and BSD, and then we can measure the differences. If there is a significant performance difference either way then we can try and investigate it to see why.

Quoted 3 lines

Thanks,

Jim

This is indeed what we should do if we get no answer from the list by someone who has already done such work.

Hans

#7 lore raw
In reply to Hans Reiser
Quoted 2 lines

If I can't get information about BSD v. Linux 2.4 networking code, then reiserfs has to get ported to BSD which will be both nice and a pain to do.

I dont think raw network data helps. 2.2 and FreeBSD are basically the same speed for raw networking in the general case. So if someone was seeing real Linux/BSD differences Im concerned it might be a driver but also that it might not have been networking differences but perhaps VM or disk I/O performance. Clearly they saw something since its rather hard to mess up that kind of measuring. I wonder if it was networking though.

The extreme answer to the 2.4 networking performance is the tux specweb benchmarks but they dont answer for all cases clearly.

Alan

#9 lore raw
In reply to Todd
Quoted 13 lines · Hans Reiser
Quoted 12 lines · Nathan Dabney
Quoted 5 lines · Hans Reiser

The problem is that I really need BSD vs. Linux experiences, not Linux 2.4 vs. 2.2 experiences, because the webcache industry tends to strongly disparage Linux networking code, so much better isn't necessarily good enough.

Hans

Check with the www.swelltech.com people, they should have the info you need.

http://www.swelltech.com/pengies/joe/squidtuneup/t1.html

Quoted 8 lines
Quoted 3 lines

The above link contains some decent squid performance hints for 2.2+Squid.

-Nathan Dabney

It does not say anything about BSD vs. Linux 2.4 networking code.

If I can't get information about BSD v. Linux 2.4 networking code, then reiserfs has to get ported to BSD which will be both nice and a pain to do.

You know Hans, both Linux v2.4 and *BSD are free. Install a copy of each and run a couple of benchmarks. I seem to recall that you have a knack for running benchmarks...

You can't always rely on having others getting all the information for you.

/David Weinehall
_ _
// David Weinehall <[email protected]> /> Northern lights wander \\

//  Project MCA Linux hacker        //  Dance across the winter sky //
\>  http://www.acc.umu.se/~tao/    </   Full colour fire           </
#10 lore raw
In reply to Todd
Quoted 6 lines · Hans Reiser
Quoted 5 lines · Nathan Dabney

On Thu, Mar 01, 2001 at 07:03:31PM +0300, Hans Reiser wrote:

The above link contains some decent squid performance hints for 2.2+Squid.

Quoted 7 lines
Quoted 1 line

-Nathan Dabney

It does not say anything about BSD vs. Linux 2.4 networking code.

If I can't get information about BSD v. Linux 2.4 networking code, then reiserfs has to get ported to BSD which will be both nice and a pain to do.

Hans

Correct, it only has information which would help tune the setup you first described (the 2.2 setup for your client).

The individuals doing the technical side of that company have a high level of knowledge regarding Linux kernel issues which affect Squid performance and can /possibly/ discuss 2.4 vs. BSD issues with you if you ask.

However, for your client I believe the "technical information" would not be as useful as demonstrated performance. I can't imagine that testing squid would require that much effort when compared to porting resierfs.

It's up to you of course.

-Nathan

#18 lore raw
In reply to Hans Reiser
Quoted 4 lines · Hans Reiser

This is indeed what we should do if we get no answer from the list by someone who has already done such work.

Hans,

exactly what you want to measure? I have UP, 2way-SMP and 4way-SMP machines all of which have at least Linux+FreeBSD installed. All my tests so far (e.g. comparing NFS servers or filesystems etc) showed Linux (2.4) to be a lot faster than FreeBSD in all areas. However, to get specific answers you need to ask specific questions. Ask and you shall receive.

(things like SPEC SFS results I can't tell because it is illegal (without going through proper steps of publishing them), I shouldn't even be saying that they show Linux to be much faster :)

Regards,
Tigran

#20 lore raw
In reply to Hans Reiser

Hello!

Quoted 2 lines

They know that iMimic's polymix performance on Linux 2.2.* is half what it is on BSD.

What is "iMimic's polymix"? I am almost sure, it is simply buggy and was not _debugged_ under linux.

Alexey

#4 lore raw
In reply to Hans Reiser
Quoted 7 lines · Hans Reiser
Quoted 1 line
Quoted 1 line

They know that iMimic's polymix performance on Linux 2.2.* is half

what it is on

Quoted 3 lines
Quoted 3 lines

BSD. Has the Linux 2.4 networking code caught up to BSD?

Can I tell them not to worry about the Linux networking code

strangling their

Quoted 1 line
Quoted 1 line

webcache product's performance, or not?

Hans, if iMimic's polygraph performance is "half" on linux versus that of freebsd, then it is a sign that iMimic has some awful code and/or are doing something wrong in linux versus freebsd.

Quoted 3 lines

The problem is that I really need BSD vs. Linux experiences, not Linux 2.4 vs. 2.2 experiences, because the webcache industry tends to strongly disparage Linux networking code, so much better isn't necessarily good enough.

please stop generalizing. there is at least one vendor in the webcache industry that is more than happy with the linux networking code.

cheers,

lincoln.

#19 lore raw
In reply to Tigran Aivazian
Quoted 15 lines · Tigran Aivazian
Quoted 4 lines · Hans Reiser

This is indeed what we should do if we get no answer from the list by someone who has already done such work.

Hans,

exactly what you want to measure? I have UP, 2way-SMP and 4way-SMP machines all of which have at least Linux+FreeBSD installed. All my tests so far (e.g. comparing NFS servers or filesystems etc) showed Linux (2.4) to be a lot faster than FreeBSD in all areas. However, to get specific answers you need to ask specific questions. Ask and you shall receive.

Quoted 6 lines

(things like SPEC SFS results I can't tell because it is illegal (without going through proper steps of publishing them), I shouldn't even be saying that they show Linux to be much faster :)

Regards,
Tigran

Thanks Tigan, you helped me move the client past the Linux vs. BSD issue.

Hans

#21 lore raw
In reply to Hans Reiser

Hello Hans,

Thursday, March 01, 2001, 7:26:20 AM, you wrote:

HR> I have a client that wants to implement a webcache, but is very leery of HR> implementing it on Linux rather than BSD.

HR> They know that iMimic's polymix performance on Linux 2.2.* is half what it is on HR> BSD. Has the Linux 2.4 networking code caught up to BSD?

HR> Can I tell them not to worry about the Linux networking code strangling their HR> webcache product's performance, or not?

HR> Hans
HR> -
HR> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in HR> the body of a message to [email protected]
HR> More majordomo info at http://vger.kernel.org/majordomo-info.html
HR> Please read the FAQ at http://www.tux.org/lkml/

It is not only related to TCP/IP performance. it is related to whole OS performance. especially performance of file system and stablity, network driver performance etc.
FreeBSD with softupdates turned on seems horrible fast and stable. but Linux 2.4 is horrible fast in TCP/IP too. diffcult to compare between in Linux and FreeBSD. don't do such stupid thing. you'll never get a correct result.

Signature · 2 lines
Best regards,
David Xu
#8 lore raw
In reply to Alan Cox

<snip stuff about someone using linux for a web cache>

Quoted 2 lines · Alan Cox

The extreme answer to the 2.4 networking performance is the tux specweb benchmarks but they dont answer for all cases clearly.

However, I think you've hit the nail on the head here; much of tux is just general-purpose network file-blasting. The right hacker could turn it into the fastest web-cache on the planet with the right modules. I believe Ingo already did a basic ftp server based on tux, just to demonstrate this generality.

Ingo? Am I crazy or enlightened?

regards,
David

#11 lore raw
In reply to Todd
Quoted 3 lines · [email protected] (Hans Reiser) writes:

If I can't get information about BSD v. Linux 2.4 networking code, then reiserfs has to get ported to BSD which will be both nice and a pain to do.

So we would get dual-licensed ReiserFS (BSD and GPL)?

Are you aware of the legal implications, making your currently GPL-only code BSD-licensed (status of third party patches for the GPL code and so on)?

Regards
Henning

Signature · 5 lines
Dipl.-Inf. (Univ.) Henning P. Schmiedehausen       -- Geschaeftsfuehrer
INTERMETA - Gesellschaft fuer Mehrwertdienste mbH     [email protected]

Am Schwabachgrund 22  Fon.: 09131 / 50654-0   [email protected]
D-91054 Buckenhof     Fax.: 09131 / 50654-20   
#13 lore raw
In reply to Todd
Quoted 10 lines · Henning P. Schmiedehausen
Quoted 3 lines · [email protected] (Hans Reiser) writes:

If I can't get information about BSD v. Linux 2.4 networking code, then reiserfs has to get ported to BSD which will be both nice and a pain to do.

So we would get dual-licensed ReiserFS (BSD and GPL)?

Are you aware of the legal implications, making your currently GPL-only code BSD-licensed (status of third party patches for the GPL code and so on)?

There would be no reason to BSD licence ReiserFS.. The intent of the BSD licence is to let anyone who wants to lock it up with more restrictive licences do so, and if the result is more popular.. take over control of the software.

So Hans could easily release a GPLed copy of FreeBSD with reiserfs. This type of activity is encouraged by the BSD people.

#12 lore raw
In reply to Henning P. Schmiedehausen
Quoted 5 lines

So we would get dual-licensed ReiserFS (BSD and GPL)?

Are you aware of the legal implications, making your currently GPL-only code BSD-licensed (status of third party patches for the GPL code and so on)?

Read Hans licensing. He's been very careful both to make that clear and cover it.

#14 lore raw
In reply to Gregory Maxwell

Hello Gregory,

Friday, March 02, 2001, 9:00:07 PM, you wrote:

Quoted 10 lines · Henning P. Schmiedehausen
Quoted 10 lines
Quoted 3 lines · [email protected] (Hans Reiser) writes:

If I can't get information about BSD v. Linux 2.4 networking code, then reiserfs has to get ported to BSD which will be both nice and a pain to do.

So we would get dual-licensed ReiserFS (BSD and GPL)?

Are you aware of the legal implications, making your currently GPL-only code BSD-licensed (status of third party patches for the GPL code and so on)?

GM> There would be no reason to BSD licence ReiserFS.. The intent of the BSD GM> licence is to let anyone who wants to lock it up with more restrictive GM> licences do so, and if the result is more popular.. take over control of the GM> software.

GM> So Hans could easily release a GPLed copy of FreeBSD with reiserfs. This GM> type of activity is encouraged by the BSD people.
GM> -
GM> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in GM> the body of a message to [email protected]
GM> More majordomo info at http://vger.kernel.org/majordomo-info.html
GM> Please read the FAQ at http://www.tux.org/lkml/

Yes, I have never heard a version -- FreeLinux. Linux is not free. FreeBSD is true free.

Regards,
linuxjob mailto:[email protected]