Re: Slow rsh performance linux->solaris

[George Bonser <grep@shorelink.com>]

On 9 Jul 1999, Morgan Fletcher wrote:

> I posted previously about slow rdump times from linux to solaris. I
> later tried dumping a local partition to another partition on the same
> linux box, and it was blazing fast with no errors. So then I just
> tried sending packets from linux to (a pipe to rsh to) solaris and
> timing it. In this environment, linux is butt-slow when compared to
> other unices. All machines are on the same 10base-t network. Check it
> out:
> 
>   linux->solaris
>   time dd if=/dev/zero count=16384 ibs=1k | rsh ale dd of=/dev/null obs=1k
>   16384+0 records in
>   32768+0 records out
>   24459+10016 records in
>   16384+0 records out
> 
>   real    1m27.993s
>   user    0m0.290s
>   sys     0m0.790s
> 

Just a quick test on my systems ... the linux box:

uname -a
Linux calvin 2.2.10-PRE2 #1 SMP Wed Jun 2 15:17:11 PDT 1999 i586 unknown

using Debian dump package: 0.4b4-13

on potato

and the solaris:

SunOS raps 5.6 Generic_105181-05 sun4u sparc SUNW,Ultra-5_10

both on the local net here:

$ time dd if=/dev/zero count=16384 bs=1000 | rsh -l grep raps dd
of=/dev/null bs=1000
16384+0 records in
16384+0 records out
11321+11298 records in
11321+11298 records out

real    0m14.819s
user    0m0.070s
sys     0m0.980s


Which is slightly better than your solaris->solaris

so my guess is that whatever the problem is, it is fixed in a newer
version of one of the following:

linux kernel
netstd package