The idea
One shared wire, several stations. If two of them transmit into it at the same moment, the two signals blur together at the receiver and neither one survives intact.
There are three broad ways to stop that happening, and every scheme this unit covers is a variant of one of them: cut the channel into pieces beforehand, let stations compete for it and recover from the occasional collision, or hand out permission to one station at a time so a collision can never happen at all.
The three families
How it works
Partitioning, random access, controlled access
Channel partitioning divides the medium into smaller pieces — separate channels — and gives each station its own piece. Nobody competes for anything, because nobody shares anything.
Random access gives every station the full channel bandwidth and lets them compete for it. CSMA/CD and CSMA/CA, the next two topics, are both random access: a station transmits when it believes the channel is free, and the protocol’s job is to handle the case where two stations guess wrong together.
Controlled access avoids the guess entirely. A station only transmits when something tells it to — a token, a reservation, a poll. Three forms of this appear in the lecture: reservation, polling, and token passing.
Token passing
A token frame circles a logical ring, moving from station to station. A station holding the token with nothing to send just repeats it on to the next station. A station that does have a frame ready retains the token, sends its data, and waits.
The receiving station replies with an ACK. Only once the sender has that ACK does it release the token back onto the ring, which hands it to the next station in line.
While one station holds the token, nobody else transmits. There is no contention and no collision — the trade is that a station with nothing to send still has to wait its turn for the token before anyone downstream gets a turn either.
Reservation
Before any data moves, a station makes a reservation. A reservation frame precedes the data frames, and inside it sits a set of mini-slots, one per station. Each station writes into its own mini-slot to claim a data slot. Nobody else’s slot can conflict with it, because the layout is fixed in advance.
Polling
Polling needs a topology where one device is designated the primary and the rest are secondaries. The primary drives the whole exchange:
- Primary broadcasts SEL (select), naming the secondary it wants to talk to.
- That secondary replies with an ACK to say it is ready.
- Primary broadcasts Poll.
- The secondary responds with either data, or a NAK if it has nothing to send.
| Family | Who decides | Collision possible? |
|---|---|---|
| Channel partitioning | Fixed in advance — each station owns a slice of the channel | No — nobody shares a slice |
| Random access | Each station guesses independently | Yes — that is the trade-off |
| Controlled access | A token, reservation or poll grants permission | No — permission is exclusive |
Check yourself
A ring has five stations. The token is currently at station C, which has nothing to send. What happens next?
C repeats the token on to the next station in the ring. Nothing is transmitted, and no station is waiting on a collision that cannot happen — that is the entire point of controlled access.
In the exam
- Name the three families and one protocol from each. Channel partitioning; random access (CSMA/CD, CSMA/CA); controlled access (reservation, polling, token passing).
- Token passing has no contention and no collision. State this directly — it is the one guarantee random access cannot make.
- The polling sequence is SEL → ACK → Poll → data/NAK, in that order. Getting the order right is worth the mark on its own.
- Reservation uses mini-slots, one per station, inside a reservation frame that precedes the data.