| T.R | Title | User | Personal
Name | Date | Lines |
|---|
| 2319.1 | Works fine that we know of.... | NETCAD::BATTERSBY | | Fri May 26 1995 10:37 | 7 |
| It doesn't have any problems that I'm aware of. We tested the
EF's ability to translate Appletalk 1 & 2 AARP, and there aren't
any hidden "gotcha's" that I'm aware of. Scott Bradner, I believe
also tested these protocols, when they tested the 900EF, and didn't
find anything in this area that I'm aware of.
Bob
|
| 2319.2 | Some clarification, please? | CGOOA::PITULEY | Ain't technology wonderful? | Fri May 26 1995 12:06 | 16 |
| Some questions that need to be answered:
1) What type of computers are being used (specific models)?
2) From your original note, I assume that one of the systems in
question is connected to an ethernet and the other is connected
directly to the FDDI. Is this the case?
3) Has the customer seen a degradation or is it just not as fast as
his expectations?
The answers to these questions will help to shed light on the problem.
Brian Pituley
NPC, Calgary
|
| 2319.3 | more info | EEMELI::PPYLKKONEN | | Mon May 29 1995 07:09 | 15 |
| The test results we received from the customer:
Mac - Ethernet repeater - Mac file size transfer time
11.4Mbyte 68 s
Mac - DECswitch 900EF -Mac 11.4Mbyte 256 s
both Mac's are connected to Ethernet ports
Mac - DECswictch900EF -FDDI - SUN
11.4Mbyte 210 s
Any idea where is the problem??
regards
pekka
|
| 2319.4 | A couple of things to look at...... | CGOOA::PITULEY | Ain't technology wonderful? | Mon May 29 1995 11:52 | 10 |
| It looks to me like there are two things at work here. The first is
that the majority of your added time is from the latency that is inherent
in going across a bridge-like device (the 900EF). The second could be
IP configuration settings like window and maximum packet size. Depending
on the IP stack being used on the Macs, you may or may not be able to tune
the Macs' IP performance.
Brian Pituley
NPC, Calgary
|
| 2319.5 | Have to get more info! | EEMELI::RUOTILA | | Mon May 29 1995 12:37 | 22 |
|
RE .4
Could the latency across DECswitch 900EF be so much bigger that with
repeater?
Yes, we have to check the IP configuration. Because we do not have sniffer
trace from the site I am not 100% sure that TCP/IP is used when they
copy!
What is EtherTALK? Is there a transport layer for this protocol which
they could utilize when they copy?
Not a MAC/EtherTALK -specialist!
Regards, Risto
|
| 2319.6 | Some questions answered...... | CGOOA::PITULEY | Ain't technology wonderful? | Mon May 29 1995 13:32 | 34 |
|
A repeater is a very simple device. The packet comes in, gets re-timed
and it goes out. Most repeaters have a latency so low that it is not
measurable.
The 900EF is a store-and-forward device. Every packet that comes in is
stored in a buffer while the source and destination addresses are
checked and CRC checks and such are performed. All of this takes time.
The 900EF is called a switch which is accurate. The difference between
it and a typical bridge is that the 900EF has multiple ports and can
conduct multiple conversations at the same time where a bridge is
usually a two port device. Some people would say that the 900EF is
actually a multi-port bridge....and they would be accurate
too......Clear as mud??????? :-)
You shouldn't *need* a sniffer-like device to check the IP performance.
All of the configuration (and present settings) should be available in
the set-up windows for the IP stack.
As far as which protocol you are using, I would just about guarentee
that you are using IP if you are talking to a Sun. (The only other
possible protocol is EtherTalk and that takes an ugly little
application on the Sun....not anyone's idea of fun.)
EtherTalk is actually just AppleTalk over an ethernet network.
EtherTalk has unfortunately become a generic term used to descibe Apple
computers using ethernet. Mac's do have the ability to run dual
protocol stacks and I would suspect that that's what your customer is
doing.
Brian Pituley
NPC, Calgary
|
| 2319.7 | Yes, just wondering... | EEMELI::RUOTILA | | Tue May 30 1995 09:47 | 11 |
|
Hi Brian,
Yes, I know the basic differens between repeaters and bridges/switches.
I am just wondering the big time differens we got in our tests.
Ok, we can check window size and packet size. Brian, do you have any
recommendations what could be the values for those parameters we could
start in this case?
Thanks Risto
|
| 2319.8 | No TCP/IP! | EEMELI::RUOTILA | | Tue May 30 1995 09:58 | 14 |
|
Hi Brian,
I just got more information. They are not using TCP/IP!
You mentioned "the only other possible protocol is EtherTalk and that
takes an ugly little appication on the Sun....not anyone's idea of fun"
Yes, they are using EtherTalk and the application they run on Sun is
EtherShare-Helios if I got it right.
Brian, could you say that how this changes the situation?
Thanks, Risto
|