Ethernet

1
Ethernet
2
Review

• Media Access Control
• Broadcast media shared by all stations
• MAC is used to determine who gets the right to
  send
• Developed protocol (All contention based)
• P0. Send at will.
• P1. ALOHA.
• P2. Slotted ALOHA
• P3. Carrier Sense. (Ask which curve is for which)
• 1-persistent
• Non-persistent
• P-persistent
• P4. CSMACD

3
Contention Free?

• The contention based protocols allows contention,
  so works well under light load.
• How to design contention-free protocols?
• The key is let other people know that I am going
  to send

4
Contention Free Protocols Bitmap

• bit-map method.
• control frame contain N bits, each station send 1
  bits to indicate whether it has a frame to send
• at the end of the control frame, every station
  knows all stations that want to send, the station
  can send in order.
• example
• Performance
• d/(d1) channel utilization rate for high load.
• N bits delay for low load. (d is the frame size).

5
Contention Free Protocols Binary Count Down

• each station sends the address bits in some order
• The bits in each position from different stations
  are ORed.
• As soon as a station sees that a high-order bit
  position that is 0 is overwrite by 1, it gives
  up.
• Eventual, only one station (with largest station
  number among all the competitors) gets the
  channel.
• Performance
• channel utilization rate d/(dlog(N)) for high
  load
• log(N) bits delay for low load.
• Contention field can serve as the address field.
• This protocol assumes that delays are negligible!

6
Contention Free Protocols Token Ring

• Passes tokens among stations and only stations
  got the token is allowed to send

7
Limited Contention Protocols

• Want to be as contention-based protocols (ALOHA)
  under light load and as contention-free protocols
  under heavy load (bitmap)
• Can we achieve that?
• The trick is to dynamically control the size of
  the group that can contend for a slot
• under light load, only one group including
  everyone and everyone can try for each slot
  this is aloha
• under heavy load, the group size is small and
  each group can only try for his slot when the
  size of the group is 1, this is bitmap

8
Limited Contention Protocols

• So, how to determine the right size of the group?
• Each station monitors the load (because it is
  broadcast media).
• If there are N stations and n have frames to
  send, divide into N/n groups

9
Limited Contention Protocols

• What happens if there is a contention inside a
  group?
• Many ways to deals with it.
• One way we will talk about is to send all frames
  of this group, then move to the next group.
• Suggestions?

10
Limited Contention Protocols
0
Like a binary search. Example
2
1
3
6
4
5
D
A
B
C
E
F
G
H
Slot 0 C, E, F, H (all nodes under node 0
can try), conflict slot 1 C (all nodes under
node 1 can try), C sends slot 2 E, F, H(all
nodes under node 2 can try), conflict slot 3 E,
F (all nodes under node 5 can try),
conflict slot 4 E (all nodes under E can try),
E sends slot 5 F (all nodes under F can try), F
sends slot 6 H (all nodes under node 6 can
try), H sends.
11
Ethernet (802.3)

• 1-persistent CSMA/CD binary exponential backoff
• Carrier sense station listens to channel first
• 1-persistent If idle, station may initiate
  transmission
• Collision detection continuously monitor channel
  and if collision, abort transmission immediately,
  and wait for a random time
• binary exponential backoff (new, how to pick the
  random time)
• each time slot is 51.2 us
• first collision, retransmission interval random
  number between 0,1
• second collision, interval random number
  between 0,1,2,3
• kth collision, interval random number between
  0, 2k-1
• upper bound 1023 slots.

12
Why binary exponential backoff

• Why not pick a random number from a fixed
  interval?
• Why a fixed small interval not good?
• Why a fixed large interval not good?

13
Ethernet Frame Format
(a) DIX Ethernet, (b) IEEE 802.3
14
Minimum Frame Size

• Why a minimum frame size is needed?
• How long does it take for a station to notice a
  collision?

15
Worst case
16
Minimum Frame Size

• So, if maximum delay is t, the minimum frame size
  is 2tbit rate.
• t is about 50us.
• So the minimum frame size of 10M Ethernet is 512
  bits.
• What if the speed goes up?

17
Ethernet Performance

• Suppose there are k stations. Let p be the
  probability that a station has a frame to send
  when the channel is idle. Assume it is
  independent across stations, and is independent
  for one station at different times. Find the
  average number of collisions before a frame is
  sent.
• First, the probability that one station got the
  chance to send is Akp(1-p)k-1.
• Second, maximized when p1/k. So A is bounded by
  (1-1/k)k-1.
• Third, each contention is indepedent, so average
  number of collision is 1/A, which is e when k is
  large.
• Each contention is 2t, so channel efficiency is
  P/P2et.

18
Switched Ethernet

• Stations connect to a switch using dedicated
  lines.
• Input frames are buffered.
• So no collision!

19
Ethernet

• Physical medium
• thin cable/thick cable/twisted pair/fiber
• 10Base5 500 meters thick (cable) Ethernet
  100 nodes/seg
• 10Base2 200 meters thin (cable) Ethernet
  30 nodes/seg
• 10BaseT 100 meters twist pair
  1024 nodes/seg
• 10BaseF 2000 meters fiber optics
  1024 nodes/seg
• 10Base5/10Base2, cable connected to each machine
  
• 10BaseT -- connecting to a hub
• 10BaseF -- between building Connecting

20
Ethernet

• Fast Ethernet
• Keep everything in Ethernet, make the clock
  faster 100Mbps.
• Cable
• 100Base-T4 100m category 3 UTP, 4 lines.
• 100Base-Tx 100m category 5 twisted pair
• 100Base-Fx 2000m Fiber optic