1. Lecture 7:
Transparent Bridges

2. The OSI reference model
enables any pair of systems in the
network to communicate with
each other
transmits chunks of information
across a link

3. Data Link Layer: Local Area Networks
When people use the term LAN, they may refer to any of a number of technologies that have the properties usually associated with LANs. Following are some of those properties.
• Multiple systems attached to a shared medium.
• "High" total bandwidth (the total bandwidth is shared by all the stations).
• "Low" delay.
• "Low" error rate.
• Broadcast capability, also known as multicast capability (the ability to transmit a single message and have it received by multiple recipients).
• Limited geography (several kilometers).
• Limited numbers of stations (hundreds).
• Peer relationship among attached stations (as opposed to a group of slaves with a master). In a peer relationship, all attached stations are equivalent. In a master/slave relationship, one special station, called the master, polls the slaves, giving each one a turn to transmit.

4. LAN Protocols The data link layer into two sublayers:
1. MAC (Medium Access Control) addresses issues specific to a particular type of LAN. For example, it deals with channel management algorithms such as token passing, binary backoff after collision detection, priorities, error detection, and framing.
2. LLC (Logical Link Control) defines the fields that allow multiple higher layer protocols to share the use of the data link.
There are several flavors of LLC. The following two types are in wide use.

5. LLC
a. Datagram protocol: a packet is delivered with best-effort service by the data link layer.
There is no protocol at the data link layer to alert the source as to whether the packet was successfully received.
Instead, error control of that sort, if needed, is assumed to be carried out at a higher layer.
Note that datagram doesn't actually do anything because the LAN already gives best-effort service.

6. LLC
b. Reliable connection-oriented protocol on top of the basic datagram:
in addition to the fields required by datagram, there are fields to number packets, provide acknowledgment field, and provide for differentiating data packets from control packets such as acknowledgments and resynchronization messages.
It is basically running the connection-oriented data link protocol HDLC (high-level data link control), which was designed for point-to-point links, on top of the LAN datagram-oriented protocol.

7. Terminilogy Bridge: layer 2 connectivity between LANs
Router: layer 3 connectivity between LANS
Bridge may also pass any layer 3 data, while performing functionality on layer 2.
Switch: vendor specific “fast” bridge

8. Reasons for bridges:
We would like to connect several LANs through some connecting unit – bridge, since:
Limited number of stations: each station that is attached causes increased delay even if it is not transmitting.
Limited size: the cable must be sufficiently short that when a station at one end of the cable transmits a packet of the legally minimum size, the transmitter will detect a collision.
Limited amount of traffic: In all LANs, the available bandwidth must be shared by all stations. The more stations that there are and the more stations that are attempting to transmit, the smaller the share of bandwidth for each station.
Broadcast: should be sent to all stations in the LAN

9. Main features
1. "no-frills" bridge: the promiscuous listen and the store and forward capabilities
2. "learning" bridge: the station learning cache
3. "complete" bridge: the spanning tree algorithm
Note: all three must exist in a standard IEEE802.1ad bridge.

10. IEEE 802 The purpose of that committee is to standardize LANs.
It has standardized several LANs.
802.1:
This committee deals with issues common across all 802 LANs, including addressing, management, and bridges.
802.2:
This committee defines LLC. MAC and physical layers are defined for a specific type of LAN by the committee that defines that type of LAN.
802.3:
This committee deals with the CSMA/CD (carrier sense multiple access with collision detection) LAN.
This is derived from the Ethernet, which was invented by Xerox and developed by Digital, Intel, and Xerox.
802.4:
This committee deals with the token bus LAN.
802.5:
This committee deals with the token ring LAN.

11. The no-frills bridge
The most basic form of transparent bridge is one that attaches to
two or more LANs
(each attachment to a
LAN is known as a port).
Such a bridge listens promiscuously to every packet transmitted and stores each received packet until it can be transmitted on the LANs other than the one on which it was received.

12. The no-frills bridge
The transparent bridge was developed to allow stations that were designed to operate on only a single LAN to work in a multi-LAN environment.
The stations expect to transmit a packet, exactly as they would in a single-LAN environment, and have the packet delivered. The bridge must therefore transmit the packet exactly as received. If the bridge modified the packet in any way—for example, by overwriting the source address portion of the header with its own address—then protocols in the stations might no longer work properly.
The bridge does change the delay characteristics, something that might affect protocols having tight timers that expect a single-LAN environment. However, most protocols either don't have such tight timers or can be adjusted.

13. The no-frills bridge – cont.
A no-frills bridge extends the capabilities of a LAN.
For example, in the case of (collision detection) it allows the length restriction necessitated by the hardware to be exceeded. If the box connecting the two LANs were a repeater instead of a bridge, the repeater would forward each bit as it was received, and a station's transmission on one side of the repeater could collide with a station's transmission on the other side of the repeater.
However, with a no-frills bridge, the packet is first received by the bridge and then stored, waiting for the LAN on the other side to become idle. It is therefore possible for two stations on opposite sides of the bridge to transmit simultaneously without a collision.

14. The no-frills bridge – cont.
Another example of how a no-frills bridge can extend the limits of a LAN is the ability to increase the number of stations in 802.5(token ring). In token ring, the total number of stations in the ring is limited because clock jitter accumulates at each station; with enough jitter, the phase lock loop is unable to lock.
A bridge solves this problem because it
implements a completely independent
instance of the ring MAC protocol
on each ring to which it attaches.
Each ring has an independent
token and a separate active
monitor (the station on which all the stations synchronize their clocks).

15. The no-frills bridge – cont.
The no-frills bridge does not overcome the LAN total bandwidth limit. If each of the LANs connected by the no-frills bridge has 10 Mb/sec capacity, the total bandwidth that can safely be used will still be 10 Mb.
This is because the no-frills bridge attempts to ensure that every packet transmitted on any LAN eventually winds up appearing on every LAN. Because each packet will appear on each LAN, the combined transmissions of all stations on all LANs cannot exceed 10 Mb(or whatever the speed of the LANs is), except that:
1. A temporary traffic peak could occur for a short interval, and, as long as the buffer capacity of the bridge were capable of storing the excess packets, none of the packets would get lost.
2. If the buffering capacity of the no-frills bridge were exceeded and the bridge needed to drop packets, the bridge might be lucky enough to drop packets that didn't need to be forwarded because the source and destination were on the same LAN. Therefore, the throughput could theoretically exceed 10 Mb.

16. The no-frills bridge – cont.
Therefore, in theory the total aggregate throughput could exceed the bandwidth of the LANs.
However, in practice, since the no-frills bridge cannot distinguish between packets that can safely be dropped and those that must be forwarded, if total bandwidth exceeds the LAN speed, then packets will be dropped before reaching their destination.

17. The no-frills bridge – cont.
Therefore, the next enhancement to the bridge solves the problem of allowing the bridge to intelligently choose which packets to drop and also allows the aggregate bandwidth to exceed the LAN speed:
Learning Bridge

18. “Learning” Bridge The strategy used by the bridge is as follows:
The bridge listens promiscuously, receiving every packet transmitted.
2. For each packet received, the bridge stores the address in the packet's source address field in a cache, together with the port on which the packet was received.
3. For each packet received, the bridge looks through its cache for the address listed in the packet's destination address field.

19. “Learning” Bridge
a. If the address is not found in the station cache, the bridge forwards the packet onto all interfaces except the one from which it was received.
b. If the address is found, the bridge forwards the packet only onto the interface specified in the table. If the specified interface is the one from which the packet was received, the packet is dropped (filtered).
4. The bridge ages each entry in the cache and deletes it after a period of time (a parameter known as aging time ) in which no traffic is received with that address as the source address.

20. “Learning” Bridge - example
Packet arrives
Bridge learning data
Operation
source
destination - A D
A from port 1
Forward to port 2
D from port 2
Forward to port 1 Q
Q from port 1
Do nothing

21. “Learning” Bridge – example with multiple ports
Packet arrives
Bridge learning data
Operation
source
destination - A D
A from port 1
Forward to port 2 and 3
D from port 2
Forward to port 1 Q
Q from port 1
Do nothing Z C
Z on port 3
Forward to port 1 and 2

22. “Learning” Bridge – example with multiple bridges
After learning:
This concept for any tree (loop-free) topology, but what about any topology?

23. “Learning” Bridge – example with multiple paths
A transmits a packet:
Each of the three bridges:
Receives the packet
Notes that A resides on LAN1
Queues the packet for forwarding to LAN2
Support B3 succeeds to transmit the packet first to LAN2.
B3 is transparent to B1 and B2. They see that a packet from A is received on LAN2.
Therefore B1 and B2:
Receive the packet
Note that now A resides on LAN2
Queue the packets for forwarding to LAN1
Now B2 succeeds to transmit the packet second to LAN2….
That’s how a loop is created.
Sometimes number of copies travelling is also increased – causing a storm.
That’s we design algorithms on bridges that prune the topology to be a subset
which is a spanning tree!

24. “Complete” Bridge -Spanning Tree Algorithm
The purpose of the spanning tree algorithm is to have bridges dynamically discover a subset of the topology that is loop-free (a tree) and yet has just enough connectivity so that where physically possible, there is a path between every pair of LANs (the tree is spanning).
The basic idea behind the spanning tree algorithm is that bridges transmit special messages to each other that allow them to calculate a spanning tree. (These special messages are called configuration bridge protocol data units or configuration BPDUs)

25. “Complete” Bridge -Spanning Tree Algorithm
The configuration message contains enough information so that bridges can do the following.
1. Elect a single bridge, among all the bridges on all the LANs, to be the Root Bridge.
2. Calculate the distance of the shortest path from themselves to the Root Bridge.
3. For each LAN, elect a Designated Bridge from among the bridges residing on that LAN. The elected bridge is the one closest to the Root Bridge. The Designated Bridge will forward packets from that LAN toward the Root Bridge.
4. Choose a port (known as the root port) that gives the best path from themselves to the Root Bridge.
5. Select ports to be included in the spanning tree. The ports selected will be the root port plus any ports on which "self" has been elected Designated Bridge.
Data traffic is forwarded to and from ports selected for inclusion in the spanning tree. Data traffic is discarded upon receipt and is never forwarded onto ports that are not selected for inclusion in the spanning tree.