More Token Rings; Wireless Networks

CS158a

Chris Pollett

Feb. 25, 2009

Outline

Keeping Rings connected
Token Ring Media Access Control
Token Ring Maintenance
FDDI
Resilient Packet Ring
Introduction to Wireless

Keeping Rings connected

Last day, we described the basics of how to transmit frames in a token ring.
Of course, this relied on machines being connected in a ring. What happens if one machine fails?
- If we do nothing, then it is no longer a ring and the token will fly off one end to oblivion, and no one will be able to talk.
- One solution is to connect each station to a electromechanical relay. If the station is healthy the relay stays open and the current goes through the given host. If the host dies and stops providing power to the relay, the relay closes, and the ring automatically bypasses the station. This only works for electrical cable, not fiber.
- Typically, several relays are packed into a single box called a multistation access unit (MSAU).
- A pair of connections (probably one cable but with an incoming/outgoing line) is then run from each host to this box.
- It is also possible to chain boxes.

Multistation Access Unit Image

Token Ring Media Access Control

Let's look at how a token ring works in more detail...
The network adaptor for a token ring contains a receiver and a transmitter.
Most of the time an adaptor is merely retransmitting the data that it receives.
When a station has the token (so is not retransmitting it), it is allowed to send one or most packets of data.
Each packet contains the destination address of the intended receiver; it may also contain a multi-cast address.
As the packet flows around the ring each node looks to see if it is the recipient. If so, it copies the packet to its buffer, but does not remove it from the ring.
The sending station is responsible for removing the packet from the ring.
There are two choices for how to do this: (a) send the token immediately after the last frame to send (early release); or (b) send the token after waiting for the last frame to go around the ring (delayed release). Initially, 802.5 used method (b), but now supports both.

Token Holding Time

The time each station can hold the token (and so send packets) before giving it to the next station is called the token holding time (THT).
We would like this time to be less when many people need to send.
On the other hand, to ensure good ring utilization, we want this time to be longer when only one station has stuff to send.
To address the issue of fair utilization, the token contains a 3-bit priority field.
Each device that wants to send a packet assigns a priority to that packet, and the device can only seize the token to transmit a packet if the packet's priority is at least as great as the token's.
To change the priority of the token, a station can set a 3-bit reservation field in a data packet to the priority of the packet it wants to send, provided this reservation field does not have a higher value already.
When a sender is finished sending, it is supposed to set the priority of the token to the value of reservation bits and set the reservation bits to the lowest priority.

Reliable Delivery

802.5 (Token Ring) uses a 2 bit field in the trailer of packets to determine if the packet got to its intended recipient.
The bits are called A and C and when the packet is sent these are both 0.
When the recipient sees the packet it sets A to 1. If it copies the frame to the network adaptor buffer, it sets the C bit to 1.
If, when the packet comes back to the sender, the sender sees A set to 0, it knows the recipient is absent. If A is set but C is not, then the sender will probably assume the recipient's buffer is full and try again later.

Token Ring Maintenance

Each token ring has one station designated as a monitor.
The monitor tries to keep the ring healthy. It does this by doing things like making sure the token is not lost.
A healthy monitor periodically announces its presence with a special control message.
If a station fails to see this message for some period of time, it assumes the monitor has failed, and will try to become the monitor.
A "claim token" is transmitted by the station trying to become the monitor.
If the station see its token go around the ring and come back to it, it then assumes it can be the monitor.
If someone else's, claim token arrives first, then it compares its address with the token's sender. If it is higher, it still becomes monitor; otherwise, it lets the other station be monitor.
To detect a missing token, the monitor maintains a timer, initially set to
```
NumStations×THT + RingLatency
```
If the timer expires without seeing a token a new token is inserted by the monitor.

Corrupt and Orphaned Frames and Dead Stations

The monitor checks for corrupt frames by looking for incorrect checksums or invalid formats to frames. This is important because, a sender address corruption might mean no one knows to take this packet off the ring.
The monitor takes corrupt frames off the ring and re-inserts the token.
An orphaned frame is a correct frame but whose sender has died (so no one will take it off the ring).
These are detected by the monitor by setting a monitor bit to 1 when the frame goes by the monitor. If the monitor sees a frame with a monitor bit of 1, it knows it's gone around the loop at least once and should be removed.
It is possible for a station to die in some way that the MSAU doesn't know the station is dead. i.e., some kind of software death.
Any station can send out beacon frames to try to figure out how far down the ring the error is and then tell the MSAU to bypass this station.

802.5 Frame Format

8	8	8	48	48		32	8	8
Start Delimeter	Access Control	Frame Control	Destination Address	Source Address	Body	Checksum	End Delimiter	Frame Status

The frame is sent using Manchester encoding, except start and end delimiters.

Access Control - contains the frame priority, and reservation bits.

Frame Control - says the higher level protocol.

Frame Status - has the A and C bits.

FDDI

Fiber Distributed Data Interface run over fiber not cable (CDDI uses cable)
It consists of a dual ring -- primary ring and a secondary ring.
If a station fails the secondary ring is joined with the primary ring on the end-point on either side of the failure to make a new ring.
Rather than have one monitor, each station does monitoring activities in FDDI.
Each node maintains an estimate total rotation time (TRT) -- the expected time for a token to make one loop of the ring.
If too much time elapses between tokens going by a station, the station can create a "claim" frame, which includes its estimate of the TRT.
If another station thinks the TRT is too high, it replaces this claim from with its own claim frame with its estimate of the TRT.
The claim frame also is a request to regenerate the token.
If a claim frame makes its way unchanged all the way back to the sender, then the new TRT is set to its value, and it gets to regenerate the token.
If the token arrives before the TRT has elapsed, a station is allowed to transmit for this extra time.

Resilient Packet Ring

This is a relatively new technology that has been standardized under 802.17 that is used in metropolitan area networks (MANs).
RPR consists of two counterrotating optical fiber rings.
It makes use of the bandwidth of both rings.
Unlike token ring and FDDI, it is the receiver of a frame who removes the frame from the ring.
RPR does not use tokens. It uses buffer insertion, a node can transmit its own frames whenever it has no frames to forward. If a frame arrives while it is sending, its own frame, then the node temporarily buffers that frame.
RPR support three Quality of Service classes to help address issues of things like starvation. class A -- low latency, low jitter; class B -- predicatable latency and jitter; and class C -- best effort.
To recover from failues RPR uses either wrapping (similar to FDDI) or steering -- nodes adjacent to the failure identify the failure and packets are then sent in appropriate directions to avoid the failure.
RPR was designed to run over both SONET and ETHERNET

Introduction to Wireless

We will now consider how the data link layer works in different wireless technologies.
We will consider Bluetooth, Wi-fi, Wi-MAX, and 3G cellular.
The most widely used wireless technologies have two different kinds of nodes: basestation -- which have no or limited mobility and client nodes which can move around.
One interesting aspect of wireless is that it naturally supports point to multi-point communication.
Another interesting aspect is the degree of mobility you might choose to support: no mobility - for instance, one might receive a direction transmission from the base station (early WiMAX); mobility within a base stations range - bluetooth, and mobility between base stations 3G cellular and Wi-Fi.