[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

irqtune



I'm trying to fix a serial link speed problem;

We are having speed problems over a serial PPP link, trying to realize the
115K capacity of the physical link. Using a commercial com server (Xyplex)
I get 115K each way; replacing it with an small Linux (Debian LRP) I get
about 70-90K max; and highly asymmetrical (testing via FTPs to local Suns).

We have two 486's; 32M each, high speed Uarts (16550). We get bursts of
fast throughput, so it doesn't seem physical (setserial, ...), but overall
throughput is too low.

Are there some "secrets" here that we should know? 
All routers and system used have MTU of 1500.

Also, 
is there a way to query the ppp driver for its current configuration?

We are looking for ways to analyse what is the cause of the slowdown; any
recommended analysis methods?

Thanks,
Greg Guthrie  guthrie@mum.edu  [please (also) reply via Email]


Someone sugested to use "irqtune"; but I then saw this, which implies that
that would not be needed, itis in the kernel; yes?
---------------------------------------
On Sun, 01 Sep 1996 11:12:42 +0300 Linus Torvalds (torvalds@cs.Helsinki.FI)
wrote:

> In my never-ending battle to make the kernel behave well by default without
> needing "irqtune" (which is very setup-specific, and as such not something
> the kernel can do automatically), I was thinking of doing interrupt priority
> rotations instead of the current fixed mode. 

Just to tell you that this patch gave me the same effect as irqtune did
before: full serial speed. Actually, I applied the 2.0.17 patch which
contained this small patch.
I'm glad this thing's in the kernel now.
---------------------------------------



--------------------------------------------------------
Gregory Guthrie
guthrie@mum.edu         (515)472-1125    Fax: -1103
       Computer Science Department
       College of Science and Technology
       Maharishi University of Management
       http://www.mum.edu/csdept
--------------------------------------------------------


Reply to: