:this is what i get, it's between a dell 2540 (dual PIII/900) and an Intel Sl2
:(dual PIII/1g). both are using intel's em driver.
:
:majadara> ./tbench 1 bagel
:Throughput 12.4785 MB/sec (NB=15.5981 MB/sec 124.785 MBit/sec)
:...
:Throughput 37.002 MB/sec (NB=46.2526 MB/sec 370.02 MBit/sec)
: ./
this is what i get, it's between a dell 2540 (dual PIII/900) and an Intel Sl2
(dual PIII/1g). both are using intel's em driver.
majadara> ./tbench 1 bagel
Throughput 12.4785 MB/sec (NB=15.5981 MB/sec 124.785 MBit/sec)
majadara> ./tbench 2 bagel
Throughput 18.1598 MB/sec (NB=22.6998 MB/sec 181.5
This part of the thread sounds really familiar. I recall someone coming
up with a patch for this a few weeks ago, possibly committing it to
-current. I'm too tired and it's too late, though; I'll look for it
tomorrow if Matt doesn't find the thread in the archives first.
Mike "Silby" Silbersac
:curious, as the loopback's MTU is normally 16384.
:Also, any idea on where does the 4096 limit (1460*2+1176) come from ?
:
: cheers
: luigi
It comes from the size of an mbuf, which is 2K. If you are trying to
send 4100 bytes of data what winds up happening is this:
On Sun, Dec 02, 2001 at 12:10:53AM -0800, Matthew Dillon wrote:
...
> There are still a couple of unresolved issues. I noticed that when
> connecting locally TCP is non-optimal... when sending a 4100 byte
> data block it sends two 1460 byte packets (maxseg), then one
> 1176 byte p
:nice, 950 Mbs which should be the theoretical maximum. what kind of CPUs
:do you have in there, and do you know how hard they were working?
:
These are 1.1 GHz duel Pentium III's. One of the cpu's is maxed out
at that transfer rate (this is -stable and the program is in the system
On Sun, 2 Dec 2001, Matthew Dillon wrote:
> This is connecting to inetd running a dd if=/dev/zero bs=32k on a
> machine with the rfc sysctl's turned on and 262144 byte send and
> receive buffers, without jumbo frames (my gigE switch doesn't support
> them :-( ).
nice, 950 Mbs wh
:On Sun, 2 Dec 2001, Matthew Dillon wrote:
:> Throughput 47.2446 MB/sec (NB=59.0558 MB/sec 472.446 MBit/sec) 20 procs
:>
:> It seems to max-out at around 75,000 packets per second (input + output).
:>
:> I doubt these results could be duplicated on anything but a DELL2550.
:> It ded
On Sun, 2 Dec 2001, Lamont Granquist wrote:
:
:What is the remaining bottleneck in these tests? CPU? Interrupts? What
:would you need to do to get that closer to the theoretical limit
:(something around 920 Mbs for GigE IIRC)?
Well, for one thing, I'd imagine that per-byte and per-copy overhea
On Sun, 2 Dec 2001, Matthew Dillon wrote:
> Throughput 47.2446 MB/sec (NB=59.0558 MB/sec 472.446 MBit/sec) 20 procs
>
> It seems to max-out at around 75,000 packets per second (input + output).
>
> I doubt these results could be duplicated on anything but a DELL2550.
> It dedicates
:
:Ahh, but there are patches floating around that do support zero-copy.
:Just ask Ken Merry and Drew Gallatin. I don't think they've been integrated
:due to lack of testing time, but they've existed for 2 or so years now.
:
http://people.freebsd.org/~ken/zero_copy/
--
Andrew R. Reiter
[EMAIL
On Sun, Dec 02, 2001 at 11:18:42AM -0800, Matthew Dillon wrote:
[...]
> :Does the FreeBSD tcp stack do zero copy (page flip the data to
> :userspace)? In the localhost case, it seems like there are two copies
> :to/from userspace there.
> :
> :--
> :Richard Sharpe, [EMAIL PROTECTED], LPIC-1
>
:
:OK Matt, that last patch did the trick.
:
:I am now getting 68 and 69Mb/s between my Linux system and the FreeBSD
:system.
Excellent!
:I have also tried the loopback interface, and I am getting 371Mb/s for 1 process,
:dropping to about 320Mb/s for 5.
Excellent!
:This seems like i
>>Does the FreeBSD tcp stack do zero copy (page flip the data to
>>userspace)? In the localhost case, it seems like there are two copies
>>to/from userspace there.
>
> It has the ability to do it via sendfile() and a few other mechanisms, but
>not as a normal part of typical read()/write().
A
>Does the FreeBSD tcp stack do zero copy (page flip the data to
>userspace)? In the localhost case, it seems like there are two copies
>to/from userspace there.
It has the ability to do it via sendfile() and a few other mechanisms, but
not as a normal part of typical read()/write().
-DG
Da
OK Matt, that last patch did the trick.
I am now getting 68 and 69Mb/s between my Linux system and the FreeBSD
system.
I have also tried the loopback interface, and I am getting 371Mb/s for 1 process,
dropping to about 320Mb/s for 5.
This seems like it is close to the limit for the machine I
I've fixed a couple of additional problems.
* tbench() assumes that accept() propogates the NODELAY tcp option.
It doesn't in FreeBSD. Er, it didn't in FreeBSD... my patch fixes
this.
* If the transwmitter sees a 0 window it stalls waiting for an ack.
However, if
17 matches
Mail list logo
