[rescue] u2 drive trays

Curtis H. Wilbar Jr. rescue at hawkmountain.net
Tue Aug 19 11:43:34 CDT 2003


>From: Dave McGuire <mcguire at neurotica.com>
>
>On Tuesday, August 19, 2003, at 11:29 AM, Curtis H. Wilbar Jr. wrote:
>> If the equipment supports 100baseTX full duplex though that would 
>> eliminate
>> collisions between the two points.  no ???
>
>   Well not exactly.  Yes, the collisions would be gone, but there's 
>still a shared medium in the middle: the switching fabric.  There can 
>be a lot of contention there.  Visualize nodes A, B, and C connected to 
>switched ports running full-duplex.  Nodes A and B both try to send a 
>packet to node C at the same time.  Many (most?) switches will use a 
>store-and-forward approach to address this issue.

OK.... but how is this different from FDDI.... if I have two FDDI
clients trying to throttle one FDDI server... that is still only
100Mbit of bandwidth at the server, and aggregated 200Mbit of clients...

So how does it differ ?

>
>   A fully-switched ethernet is still a shared medium, but the media 
>arbitration is handled by a switching fabric rather than detecting 
>collisions and retransmitting.  The *medium* is effectively the switch, 
>rather than the wire.

ok..

>
>   Ethernet is described as a CSMA/CD system...Carrier Sense Multiple 
>Access with Collision Detect.  If a node needs to transmit, it looks at 
>the wire to see if there's activity...if not, it goes ahead and 
>transmits.  If another node happens to do so simultaneously, you get a 
>collision...which results in garbled noise on the wire.  Since an 
>ethernet interface always sees its own packets, it knows whether 
>they've gone out ok or have been stomped on.  In the event of a 
>collision, the interface waits a random period of time and attempts to 
>retransmit.

all ethernet theory.. I'm aware of this... hence why I use switches
and detest hubs....

I was a cofounder of a regional ISP, and we witnessed during our early
growth the rapidly decreasing returns of ethernet when collisions occur...
and that is when we bought our first switch and ditched the hubs....

>
>   I don't have a background in statistics, so I don't know the correct 
>terminology here...but if you visualize this behavior over time, it 
>becomes obvious that, for a given amount of data, the more packets that 
>are being transmitted results in a higher probability for 
>collisions...even for the same amount of data.  Ethernet's tiny MTU 
>results in lots and lots of packets.

exactly what happened at the ISP I cofounded...

>
>   CSMA/CD is at the very core of Ethernet's design.  When running a 
>fully-switched network with full-duplex ports, you've modified that 
>aspect of its operation...it's effectively no longer CSMA/CD.  In fact, 
>at that point, I'd hesitate to even call it Ethernet...it's that much 
>of a change.

It's that much better IMHO.... collisions suck...

>
>   Not to be a name-dropper, but...oh, hell, I'm definitely being a 
>name-dropper.  Everyone knows that Ethernet was designed by DEC, Intel, 
>and Xerox (hence DIX).  It was later standardized by IEEE committee 
>802.3.  That committe's chairman was an old IBM programmer (he wrote a 
>chunk of what became known as OS/360) named Gerry Clancy, Jr.  He runs 
>a tiny (four people at the time) software development company in 
>Trenton, NJ called Princeton Desktop Systems, inc.  I worked for him 
>for about three years doing database programming and desktop publishing 
>stuff from about 1988.  I learned a *lot* from him.

cool...

>
>           -Dave
>
>--
>Dave McGuire                 "You don't have Vaseline in Canada?"
>St. Petersburg, FL                     -Bill Bradford
>_______________________________________________
>rescue list - http://www.sunhelp.org/mailman/listinfo/rescue

Other than MTU, I don't see a big difference between 100baseTX switched
and FDDI...  in the example you cite, how is FDDI going to do any better
than switched 100baseTX (other than an improved MTU) ?

-- Curt

Curtis Wilbar
Hawk Mountain Networks
rescue at hawkmountain.net

My e-mail is protected against viruses and spam by MailGuardian
                  http://www.mailguardian.net
          Top notch protection at unbelievable prices



More information about the rescue mailing list