| T.R | Title | User | Personal
Name | Date | Lines |
|---|
| 1264.1 | isn't this how ethernet works? | NETCAD::ROLKE | The FDDI Genome Project | Fri Feb 21 1997 08:19 | 19 |
| >Using NETBEUI........1-5% collision on a 60-70% LAN utilizacion...i.e very
>good
>
>Using TCPIP only (MS) 25% collisions on a 60-70% LAN utilizacion...i.e NOT
>VERY GOOD At ALL.
Are you saying that the same machine, same driver, same adapter, same
network gives different collision rates for different protocols?
This narrows the problem down to the protocols. There is nothing more
that the driver and adapter can do for you; they offer the protocol's
load to the network and collisions happen.
Do you have the numbers of packets transmitted by each station in the
NETBEUI and TCPIP cases? This could be a clue as to why TCPIP has more
collisions.
Regards,
Chuck
|
| 1264.2 | More data + 3com OK | NETRIX::"[email protected]" | Sunny Spain | Fri Feb 21 1997 13:41 | 36 |
| Hi Chuck, thanks for the reply,
>Are you saying that the same machine, same driver, same adapter, same
>network gives different collision rates for different protocols?
Yep, however the customer teested on Prioris ZX6200/4 and I tested with
Prioris XL590, same cards + the client also tested on another 4 DE-500-AA's he
has...same result. As regards to drivers NT4.0 native + v260 and v263 all gave
the same result.
>This narrows the problem down to the protocols. There is nothing more
>that the driver and adapter can do for you; they offer the protocol's
>load to the network and collisions happen.
The client has a applied a work-around ...install 3COM 10/100 and collision
rate (under IP or Netbuie) drops to 1-5% !...so I don't think its a generic
Microsoft TCPIP problem, otherwise I hopew we would have heard.!
>Do you have the numbers of packets transmitted by each station in the
>NETBEUI and TCPIP cases? This could be a clue as to why TCPIP has more
>collisions.
I'm not a LAN man but with a little help from my friends (NSIS) in the digital
offices we tested on an isolated LAN with a DEC Portswitch 900:-
Under Netbuie: 130000 packet txd/rxd with around 200-650 collisions
Under TCPIP: 130000 packets txd/rxd with around 29000 collisions
In both cases from server A to server B and vica versa we copied 53MB of data,
LAN utilizacion in both cases around the 60-70% mark.
I don't now have the servers to test on any more as there at ASLAN event but I
do have the cards.
Cheers John
[Posted by WWW Notes gateway]
|
| 1264.3 | Capture Effect in action? | NETCAD::ROLKE | The FDDI Genome Project | Fri Feb 21 1997 15:41 | 20 |
| There are some notes about high collision rates on networks with
few (say, two) nodes on them.
See TURRIS::digital_unix note 3056.1 for pointers to discussons of
the capture effect.
See TURRIS::digital_unix note 373.7 for adjustments you might make to
your tcpip stack to partially compensate for the effect. This note
has good explanations of what it is doing.
Does your driver have a "capture effect" kind of switch? This might make
a difference if the ethernet chip can run a different backoff algorithm.
The reason the 3Com board works differently is probably because it leaves
a larger interpacket gap between frames. This allows the second node to
start transmitting without a collision.
These are tough problems to solve!
Chuck
|
| 1264.4 | Thanks, will read up and test | NETRIX::"[email protected]" | Sunny Spain | Fri Feb 28 1997 03:49 | 9 |
| Hi Chuck,
Thanks for the reply, I will read up on said articles and hopefully test at
the client site.
Yes there are only a couple of nodes on the particular ethernet segment.
Cheers John
[Posted by WWW Notes gateway]
|