Very Computer

by **Lawrence Houst** » Sat, 07 Jan 1995 16:47:48

In moving from the Net2Debugged [1.30] (Kernel 1.0.9) to NET3.014 (Kernel
1.1.16) or NET3.017 (Kernel 1.1.59) I have noticed a sharp drop (1.5-1.6Kb/s
down to 1.2Kb/s) in my FTP Transfers over a 14.4K CSLIP Connection to second
Linux Host (Kernel 1.1.59) which is accessible via our Terminal Server!
During the same CSLIP Connection to a SUN SparcStation (on the same LAN as
the second Linux Host), I still get the 1.5-1.6Kb/s which I was previously
getting with my 1.0.9 Kernel? My MTU is set to 1500 (matching our TS) and
I am referring to transfers of compressed binary data.

In general I understood that the NET3 Code was an improvment over the NET2D
Code, so I find this drop in performance a bit surprising! Watching the
RD/SD LEDs on my USR Sportster there is some idle time with the 1.1.16/1.1.59
Kernels running, but still NOT enough to account for 25% lose in throughput.

Anybody able to suggest a cause for these inefficiencies and/or able to
provide a solution?

--

by **Stelios Parnassid** » Tue, 10 Jan 1995 08:11:29

: In moving from the Net2Debugged [1.30] (Kernel 1.0.9) to NET3.014 (Kernel
: 1.1.16) or NET3.017 (Kernel 1.1.59) I have noticed a sharp drop (1.5-1.6Kb/s
: down to 1.2Kb/s) in my FTP Transfers over a 14.4K CSLIP Connection to second
: Linux Host (Kernel 1.1.59) which is accessible via our Terminal Server!
: During the same CSLIP Connection to a SUN SparcStation (on the same LAN as
: the second Linux Host), I still get the 1.5-1.6Kb/s which I was previously
: getting with my 1.0.9 Kernel? My MTU is set to 1500 (matching our TS) and
: I am referring to transfers of compressed binary data.

I have exactly the same phaenomen. From 1.1.59 up to 1.1.75. I always thought
it's crowdy traffic on the Intenet that causes it.

sp

--
.................................................
Stelios Parnassidis :
Kapuzinerstr. 8 :
80337 Muenchen : Tel. +49-089-7211505

by **Henny Wilbri** » Tue, 10 Jan 1995 17:45:37

>: In moving from the Net2Debugged [1.30] (Kernel 1.0.9) to NET3.014 (Kernel
>: 1.1.16) or NET3.017 (Kernel 1.1.59) I have noticed a sharp drop (1.5-1.6Kb/s
>: down to 1.2Kb/s) in my FTP Transfers over a 14.4K CSLIP Connection to second
>: Linux Host (Kernel 1.1.59) which is accessible via our Terminal Server!
>: During the same CSLIP Connection to a SUN SparcStation (on the same LAN as
>: the second Linux Host), I still get the 1.5-1.6Kb/s which I was previously
>: getting with my 1.0.9 Kernel? My MTU is set to 1500 (matching our TS) and
>: I am referring to transfers of compressed binary data.
>I have exactly the same phaenomen. From 1.1.59 up to 1.1.75. I always thought
>it's crowdy traffic on the Intenet that causes it.
>sp
>--
>.................................................
>Stelios Parnassidis :
>Kapuzinerstr. 8 :
>80337 Muenchen : Tel. +49-089-7211505

I have also noticed this, both with cslip and ppp. But the really
weird thing is that the first cslip or ppp session after a HARD reboot
gives me 1.6 Kb/s. Every subsequent session gives me something in the
order of 1.3 Kb/s. A soft reboot does not help. This just doesn't make
sense. Has anybody else seen this strange behavior?

Henny Wilbrink

by **Lawrence Houst** » Wed, 11 Jan 1995 02:35:48

: : In moving from the Net2Debugged [1.30] (Kernel 1.0.9) to NET3.014 (Kernel
: : 1.1.16) or NET3.017 (Kernel 1.1.59) I have noticed a sharp drop (1.5-1.6Kb/s
: : down to 1.2Kb/s) in my FTP Transfers over a 14.4K CSLIP Connection to second
: : Linux Host (Kernel 1.1.59) which is accessible via our Terminal Server!
: : During the same CSLIP Connection to a SUN SparcStation (on the same LAN as
: : the second Linux Host), I still get the 1.5-1.6Kb/s which I was previously
: : getting with my 1.0.9 Kernel? My MTU is set to 1500 (matching our TS) and
: : I am referring to transfers of compressed binary data.

: I have exactly the same phaenomen. From 1.1.59 up to 1.1.75. I always thought
: it's crowdy traffic on the Intenet that causes it.

Stelios:

I observe the same drop in transfer rates whether it is 15:00 or 03:00! In
addition the "second" Linux 1.1.59 Host is "local" to our institution, .ie
hardwired LANs between the Terminal Server and the other PC (no WAN delays
to cause problems). Also noticed a similar "drop" with another "local" host
running HP-UX, but all the SunOS Based Hosts were all in the 1.6 Kb/s range?
Hence I do NOT believe it is rated to net traffic.

by **Duane Ca** » Sun, 15 Jan 1995 01:05:52

>>I have exactly the same phaenomen. From 1.1.59 up to 1.1.75. I always thought
>>it's crowdy traffic on the Intenet that causes it.

>I have also noticed this, both with cslip and ppp. But the really
>weird thing is that the first cslip or ppp session after a HARD reboot
>gives me 1.6 Kb/s. Every subsequent session gives me something in the
>order of 1.3 Kb/s. A soft reboot does not help. This just doesn't make
>sense. Has anybody else seen this strange behavior?

I also have noticed a drop in FTP speeds between 1.0.9 and 1.1.59 (complete
SLACKWARE re-install).

I noticed this both with my PPP connection to the provider, and with a
38400 hard-wire link to my Sparc-station. I have not messed with the
MTU setting (yet)...

Duane Cady

Very Computer

FTP Transfer rates declined with NET3 ???

FTP Transfer rates declined with NET3 ???

FTP Transfer rates declined with NET3 ???

FTP Transfer rates declined with NET3 ???

FTP Transfer rates declined with NET3 ???

FTP Transfer rates declined with NET3 ???