🚀 go-pugleaf

In article <1995Feb13.020829.14593@rd.scitec.com.au> johns@rd.scitec.com.au (John Saunders) writes:
>I've heard this so many times that it must be true. Does anybody know where
>the bottle neck is. i.e. is it
>
>1) Linux networking.
Not especially except when you are forced to use 8K block sizes to talk to
BSDish systems.
>2) Linux NFS client.
Very much so (actually much of this is the buffer cache). To be worked on
for 1.3
>3) NFS server.
The Linux nfs server isnt especially quick and handles horribly with lots
of clients. Known problem people are working on. BSD kernel based servers
have other problems with needing 4K block sizes to work efficiently. Linux
wants 1K client sizes for best results and together you get the worst of
both worlds.
>4) Ethernet drivers.
Definitely not.
>
>Because if it's the NFS server then using the BSD one should provide the
>same speed. But I have noticed ftp speed under Linux isn't as fast as what
>I expected. I get about 200K/s to 300K/s over unloaded ethernet bus. So
>maybe the TCP/IP stack under Linux is a bit slower than FreeBSD.

A bit. I get about 750K/second over an unloaded ethernet but Im using a
rather hot ISA card and a good SCSI controller. Primarily BSD has tweaks
that optimize performance for one way data transfers (eg ftp) and for
cases where a continual stream of data hits the same socket. Other areas
of the BSD stack such as mbufs are a little flawed in concept but the good
outweighs the bad. Also check your IDE drive. Linux drives IDE block
transfers with interrupts off by default so that some older dodgy drives
that can't cope with interruption of data flow work. I don't know which way
FreeBSD defaults.


Alan
-- 
  ..-----------,,----------------------------,,----------------------------,,
 // Alan Cox  //  iialan@www.linux.org.uk   //  GW4PTS@GB7SWN.#45.GBR.EU  //
 ``----------'`--[Anti Kibozing Signature]-'`----------------------------''
One two three: Kibo, Lawyer, Refugee :: Green card, Compaq come read me...