Presentation is loading. Please wait.

Presentation is loading. Please wait.

Source Routing Bridges

Similar presentations


Presentation on theme: "Source Routing Bridges"— Presentation transcript:

1 Source Routing Bridges
Lecture 7: Source Routing Bridges

2 Transparent Bridge vs. Source Routing Bridge
All standard bridges must support transparent bridging Source routing is an optional additional feature in a bridge. A bridge that does source routing in addition to transparent bridging is known as an SRT (source routing transparent) bridge.

3 Principle of work A packet header contains a route
That route is inserted by the source end station. For the source end station to know a route to the destination end station, it must discover a route by transmitting a special kind of packet that replicates itself as it reaches route choices sending a copy over each possible path. Each copy collects a diary of its travels so that when the copies reach the destination station, a route can be selected. When a station discovers a route to another station, it caches the route so that it can be used for subsequent packets to the same destination.

4 Packet structure A source routing packet requires additional information in the header, contained in a field known as the RI ("routing information") field, in order to distinguish packets that have the additional fields from those that don't since it could be unfortunate if a user's data were interpreted as source routing header information. The multicast bit in the src address is not useful. Nobody should be transmitting a packet from a multicast address. If that bit in the source address is 0, the assumption is that the packet is ordinary; if that bit is set, the assumption is that the information following the regular data link layer header should be interpreted as a source routing header.

5 The RI field contains the additional source routing information
The RI field contains the additional source routing information. It consists of the following fields. • Type (3 bits): One of the following: a. Specifically routed (the route appears in the header) b. All paths explorer (the route gets collected as packet copies traverse the network) c. Spanning tree explorer (the route gets collected as packet copies traverse the network, just as with the all paths explorer, but the packet travels only along branches in the spanning tree) • Length (5 bits): Specifies the number of bytes in the RI field • Direction (1 bit): Specifies whether the route should be traversed from right to left or vice versa • Largest frame (3 bits): A value representing one of a few popular packet sizes (516, 1500, 2052, 4472, 8144, 11407, 17800, 65535) • Route: A sequence of 2-byte-long fields, called route designators, each of which consists of a 12-bit LAN number followed by a 4-bit bridge number (see Figure 4.4).

6 Route Each bridge is further assigned a 4-bit number to distinguish it from other bridges that connect the same pair of LANs. Each bridge must be configured with the LAN numbers for each port plus a bridge number. A route is really an alternating sequence of LAN and bridge numbers, always starting and ending with a LAN number. (LAN, bridge, LAN, bridge, LAN, bridge, LAN). The final route designator's 4 bits indicating "bridge" are irrelevant.

7 Bridge algorithm: Transparent packets (no RI field)
A pure source routing bridge explicitly ignores packets that do not have an RI field. To avoid confusion in case both source routing and transparent bridges attach to the same LAN, the transparent bridge standard requires that a transparent bridge ignore packets that have the RI field. So transparent packets are handled by the transparent bridges, and packets with the RI field are handled by the source routing bridges.

8 Bridge algorithm: specifically routed packets
Assume that bridge B receives a specifically routed packet on a port P1 attached to a LAN that B knows as "X." B checks the direction bit in the header and scans the route in the appropriate direction searching for LAN X. B forwards the packet onto port P2 if all of the following are true. 1. X appears in the route. Now assume that the bridge number following X is some number, Bn, and the LAN number following X is some number, Y. 2. The LAN number that B associates with port P2 is Y. 3. B's bridge number with respect to the pair of ports (P1, P2) is Bn. 4. Y does not appear anywhere else in the route. (Otherwise, a packet could be created that would loop)

9 Bridge algorithm: All Paths Explorer Packets
Assume that bridge B receives an all paths explorer packet on a port P1 attached to a LAN that B knows as "X." B must first check to see whether the packet has accumulated a route. If the RI field contains only the packet type and flags but no actual hops yet, B does the following for every port P other than port P1. 1. B initializes the route to "X, Bn, Y," where Y is the LAN number with which B has been configured for P and Bn is B's bridge number. 2. B adjusts the value of the largest frame field to be the minimum of the value the end system placed in that field, the value configured into B as the largest frame possible on P1, and the value configured into B as the largest frame possible on P. 3. B recalculates the CRC (cyclic redundancy check) on the packet because the packet has been modified. 4. B transmits the packet onto P. If the all paths explorer packet has been through other bridges, B does the following for every port P except the port (P1) from which it received the all paths explorer packet. B forwards it onto port P (for which B has LAN number Y configured) if the following are true. 1. The final hop in the collected route is X. (If it isn't, B drops the packet and logs an error.) 2. Y does not appear anywhere in the route collected so far. Before forwarding the packet onto port P, B does the following. 1. B adjusts the length of the RI field by adding 2 to the length field in the RI header. 2. B writes its bridge number into the bottom 4 bits of the route designator field that contains LAN number X. 3. B appends Y as the next LAN number. 4. B adjusts the largest frame field if the value configured into B as the largest frame possible on P is smaller than the value currently indicated in the largest frame field. 5. B recalculates the CRC because the packet has been modified. If the route in the received all paths explorer packet is full, the packet is discarded.

10 Bridge algorithm: Spanning Tree Explorer Packets
One obvious use of spanning tree explorer frames is in the transmission of multicast packets. Multicast packets cannot be specifically routed because they need to reach multiple destinations. It is undesirable for them to be sent via all paths explorer because multiple copies of each multicast packet would be delivered to each LAN. Except for multicast situations, the standard has never been clear about when this type of packet is used. There are many possible scenarios for how an end system discovers a route. In one scenario, the conversation initiator sends the first packet of a conversation via spanning tree explorer and the destination replies with an all paths explorer. It is the source rather than the destination that chooses the route. To support this type of packet, source routing bridges run the spanning tree algorithm. The spanning tree that results is used only for spanning tree explorer packets. A bridge handles a spanning tree explorer packet almost the same way as it handles an all paths explorer. There are only two differences. 1. The bridge does not check whether the output LAN already appears in the collected route. 2. The bridge accepts a spanning tree explorer packet only if it arrives on a port in the spanning tree and forwards it only to other ports in the spanning tree.

11 End-system algorithm An end system keeps a cache of routes for destinations with which it is currently having conversations. If no route for a particular destination is in the cache, the end system can employ a protocol to find a route or set of routes. If a route in the cache no longer works and the end system discovers this, the end system can either attempt to find another route or use one of the alternate routes it has stored for that destination. An end station that is capable of source routing must be able to communicate with an end station that uses only transparent packets Even two end stations are capable of communicating via source routing, if the only paths between them involve at least one transparent-only bridge, the two stations must in that case communicate with transparent packets.

12 Source Routing vs. Transparent Bridging
Bandwidth overhead The number of paths in a network is exponential. Theoretically, an all paths explorer packet will generate a copy of itself for each path through the network. Thus, theoretically, the number of copies of one all paths explorer packet that will be generated is exponential. Source routing bridges also use extra bandwidth because the headers must be larger Transparent bridges do not make optimal use of bandwidth either; they waste bandwidth by using only a subset of the topology, by not using optimal routes, and by forwarding packets needlessly before cache entries are established. But their bandwidth waste is not nearly as spectacular as with source routing.

13 Source Routing vs. Transparent Bridging
2. Ease of Configuration Transparent bridges are truly plug and play. Although there are plenty of parameters a network manager can use to fine-tune performance in a specific situation, there is no need to modify any parameters unless desired. Source routing requires that every LAN be assigned a number and that every bridge be configured both with the LAN number for each of its ports and with a 4-bit bridge number for each pair of LANs it connects (although establishing restrictions on the topologies might enable a bridge to have only a single 4-bit number that would be unique for each pair of LANs it connects). Misconfiguration can cause problems such as loops or severe duplication of data packets.

14 Source Routing vs. Transparent Bridging
3. Universality Source routing requires support in the endnodes. Transparent bridges are not completely transparent, especially when used to connect different types of LANs. The packet size problem can be solved if the network manager configures all the endnodes to use the smallest maximum packet size allowed by any of the LANs. The source routing protocol does a fair job of solving the packet size problem because when a route is discovered, the maximum packet size on the path is also discovered. It is not necessarily easy to use the information about the maximum packet size on a route. It requires that the low-layer source routing process be able to share this information with the transport layer process, which makes decisions about packet sizes. If a source route changes in the middle of a transport connection, many end station implementations would not be able to change the packet size.

15 Source Routing vs. Transparent Bridging
4. Cost and Performance of Bridges The number of endnodes in a network is vastly larger than the number of bridges, so an increase in a bridge's cost in order to ensure its adequate performance would likely be negligible compared with the cost of the network as a whole. Because source routing requires more complicated endnodes, it might increase the cost of the endnodes. When the additional endnode cost is multiplied by the number of endnodes, the resulting figure would probably far outweigh any increase in the cost of the bridge. However, implementations of transparent bridges proved that transparent bridges with adequate performance could be built at a reasonable cost.

16 Exercise 6 (for 3 weeks) Explain the functionality that should be performed by the SR-TB bridge. 2. Suggest a bridge algorithm.


Download ppt "Source Routing Bridges"

Similar presentations


Ads by Google