At 07:24 AM 2/8/02, Surya Prakash wrote:
>Right. More light. Collisions are possible only on technologies like
>CSMACD, CSMACA TR. In these technologies there is a contest for
>bandwidth.

Your answer is of course right, that serial links don't encounter 
collisions. Collisions are an issue on networks where multiple devices 
contend for the shared transmission medium.

I don't know what you mean by CSMACA TR (hopefully not Token Ring which 
uses token passing, not CSMA! ;-) Multiple stations don't send at once in 
Token Ring. Instead the token passes from station to station with a bit 
that says whether the token is "free" or not, which means a station can send.

But a couple good examples of CSMA/CA are Local Talk (remember that?) and 
802.11B wireless. Despite the A standing for Avoidance in CSMA/CA, 
collisions do occur in those technologies.

Media Access Control on 802.11B wireless networks is quite interesting. 
 From what I understand, stations sense and then wait a random amount of 
time before sending. When sensing, a station can take into account how long 
another station will be sending because the station includes a duration 
value. Check this AiroPeek output. See the Duration field? Cool, eh?

802.11 MAC Header
        Version:              0
        Type:                 %00
        Subtype:              %0101
        To DS:                0
        From DS:              0
        More Frag.:           0
        Retry:                0
        Power Mgmt:           0
        More Data:            0
        WEP:                  0
        Order:                0
        Duration:             218  Microseconds
        Destination:          00:A0:F8:9B:B9:AA  Client B9:AA
        Source:               00:A0:F8:8B:20:1F  AP 20:1F
        BSSID:                00:A0:F8:8B:20:1F  AP 20:1F
        Seq. Number:          3095
        Frag. Number:         0

802.11B differs from other CSMA methods in a few other ways also. There's 
no collision detection. Instead, a station ACKs (at the data-link layer). 
That way the sender knows that its frame got there and no collision occurred.

802.11B also provides a fragmentation service. This has to do with 
collision avoidance (not MTU as in the IP world). The idea is to reduce 
collisions by making sure that nobody hogs the medium for too long. If a 
station were to hog the medium for a long time, the odds of two or more 
stations trying to send as soon as the original one finished increase, as 
do the collisions.

The fragmentation threshold (a configurable parameter in an AP and wireless 
NIC) can be lowered, causing large Ethernet packets to be broken into 
smaller groups of 802.11 packets. Because each packet is smaller, it may 
allow other stations to jump in to the wireless medium and take turns 
without as many collisions. The fragmentation threshold defines the 
threshold above which the MAC layer will fragment packets into a series of 
smaller packets.

802.11B also supports a slightly different method for avoiding collisions, 
which is through the use of RTS and CTS packets. (That's what LocalTalk 
does too, by the way).


Someone else might know about the Media Access Control used on cable 
modems. It may be CSMA also?


This has nothing to do with the phantom collision report on Cisco serial 
interfaces (which is just one of those silly gotchas on Cisco tests). On a 
serial link, each side of the point-to-point link has its own dedicated 
transmit circuit. Sharing isn't happening, so collisions don't happen either.

Priscilla





>-----Original Message-----
>From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED]] On Behalf Of
>Sean Knox
>Sent: Friday, February 08, 2002 10:35 AM
>To: [EMAIL PROTECTED]
>Subject: RE: collissions on serial line? [7:34816]
>
>
>A serial line is a point-to-point link. Collisions are only possible on
>a shared medium, such as ethernet.
>
>-----Original Message-----
>From: somera cecilia [mailto:[EMAIL PROTECTED]]
>Sent: Thursday, February 07, 2002 7:46 PM
>To: [EMAIL PROTECTED]
>Subject: collissions on serial line? [7:34816]
>
>
>Folks, I've been searching CCO but cannot find answers to this. Is it
>possible to get collissions in serial lines? If there are, what could
>cause this?
>_________________________________________________________


________________________

Priscilla Oppenheimer
http://www.priscilla.com




Message Posted at:
http://www.groupstudy.com/form/read.php?f=7&i=34896&t=34896
--------------------------------------------------
FAQ, list archives, and subscription info: http://www.groupstudy.com/list/cisco.html
Report misconduct and Nondisclosure violations to [EMAIL PROTECTED]

Reply via email to