© Janice Regan, CMPT 128, CMPT 371 Data Communications and Networking BGP, Flooding, Multicast routing

Janice Regan © Oct Some ASs A1A1 A2A2 A4A4 A3A3 C2 C1 B2 C4 B6 B1 B3 B5 B4 C5 AS A AS B AS C C3 IRP B IRP C IRP A Gateway router

Janice Regan © Oct ERP : Approach?  Link-state and distance-vector not effective for exterior router protocol  Link state requires flooding of link state information, for a large net this is not practical  Distance-vector requires all routers share common distance metric, different ASs may use different metrics  ASs may have different priorities such as restrictions that prohibit use of certain other AS, Distance-vector gives no information about ASs visited on route (policy routing)

Janice Regan © Oct Path Vector Approach  No path cost information used  Each block of information lists all ASs visited on a route  Allows the receiver to know the source for each path and whether the path originates in the local AS (coming from IRP or ERP)  Can be used to check for loops (any node appearing more than once)  Enables router to perform policy routing based on Avoiding transiting a particular AS link speed, capacity, tendency to become congested, overall quality of operation, security minimizing number of transit ASs

Janice Regan © Oct Border Gateway Protocol  BGP is the preferred ERP for or use with TCP/IP internets  BGP messages are sent over reliable TCP connections between gateway routers, a BGP session includes all messages sent through one of these TCP connections.  4 message types: Open, Update, Keep Alive, Notification  Gateway routers running BGP are know as BGP peers  peers may be in different ASs, eBGP or external BGP session  Peers may be in the same AS, iBGP or internal BGP session

Janice Regan © Oct Some ASs A1A1 A2A2 A4A4 A3 C2 C1 B2 C4 B6 B1 B3 B5 B4 C5 AS A AS B AS C C3 IRP B IRP C IRP A Gateway router

eBGP and iBGP  Consider the previous slide  eBGP could be used to transfer path information between gateway routers A3 and C5 and between routers C2 and B3.  AS B has more than one gateway router  AS B uses iBGP to transfer information between gateway routers in AS B Janice Regan © Oct

7 Border Gateway Protocol  Procedures that are part of BGP  Neighbor acquisition: Determine if a router physically connected to this router is willing to be a neighbor and Initiate neighbor relationship, negotiating parameters  Neighbor reachability: maintain neighbor relationship  Network reachability: build/maintain routing database

Janice Regan © Oct BGP: neighbor acquisition  Open TCP connection between a pair of connected (neighbor) routers  Each of the pair of routers sends an Open message  Includes proposed hold time, senders AS #, identifier  Identifier is an IP address uniquely identifying the sender  Each of the pair of routers receives the others Open message, If it wishes to be a neighbor it will respond to the Open message with a Keep Alive message (like an ACK) and  Select the minimum of local/received hold time, to give time between subsequent Keep Alive and/or Update messages

Janice Regan © Oct BGP: reachability  If no Keep Alive or Update message is received during the agreed upon hold time the connection is terminated.  If a neighbor wishes to continue the neighbor relationship but has no routing update to send it will send a Keep Alive message once per hold interval  Each BGP router maintains a database of reachable networks. When a change is made to this database, that is when new or updated routing information is available the router will send an Update Message  Update includes, a list of routes being withdrawn and information about new routes to be added  Each update message may contain multiple paths to add but includes one set of path information for all these paths

Janice Regan © Oct Routing Information: Paths  Each path consists of a list of ASs visited and a list of networks (CIDR network address/prefix) reachable through the gateway routers in each of the ASs visited.  When a BGP peer learns of a new path it will create an entry in its routing table for that path. Once it learns that path it can use the path.  A BGP peer may choose to advertise a path. An advertised route can be used by your neighbors to reach all the networks in the path  Advertised networks may be aggregated and advertised as one network (may be one network in the routing table)

Janice Regan © Oct  Customers of provider with AS T, have been allocated addresses that form AS X and Y AS: T / /23 AS: X /24 AS: Y /24 Example: AS path construction A B C D E To AS Z

Janice Regan © Oct Announcing paths  Want to send information about the path to AS T and the path through AS T to routers outside AS T to build a path from outside AS T to AS T  Consider a AS Z connected to T by a point to point connection from router C to router X in AS Z ( a neighbor of T)  Simplest way to advertise the networks reached in and through T is to announce three paths (1 to each AS) Path 1: “T,” reaches /23 Path 2: “T,X, “ reaches /24 Path 3: “T,Y,” reaches /24

Janice Regan © Oct BGP routing  Each BGP peer has it own import policy  Can choose to accept a new route or ignore it  If it accepts the route it can choose if it will advertise that route (make itself an intermediate step on a route from an external source to an external receiver)  Each time a BGP peer chooses to accept and advertise a new route it will append its own locally accessible networks to the path.  Before appending it will check that those networks are not already a part of the path. If they are a circular route has been detected and the route must be dropped.

Broadcast and multicast routing  We will consider 3 approaches  Uncontrolled Flooding  Sequence number Controlled Flooding  Reverse Path Forwarding  Spanning Tree broadcast Janice Regan © Oct

Janice Regan © Oct Uncontrolled Flooding  Requires no information about the network  A packet that is being sent from A to B is  Sent to all the nearest neighbors of A  Each neighbor receives the packet, then transmits the packet to all it’s own nearest neighbors, except the one it received the packet from  The packet takes all possible paths through network to B  Multiple copies of the packet will arrive at B, the first copy of the packet will arrive along the minimum cost path through the network.

Janice Regan © Oct Flooding: Example  HOP 1: The source station broadcasts the packet to all adjacent nodes. I I J L K E FH G B D C A

Janice Regan © Oct Flooding: Example  HOP 2: The receiving stations broadcast the packet to all their own nearest neighbors. The receiving stations do not broadcast back to the station they received the message from. I J L K E FH G B D C A

Janice Regan © Oct Flooding: Example  HOP 2: follow the packets I I J L K E FH G B D C A

Janice Regan © Oct Flooding: Example  HOP 3: The receiving stations for hop 2 broadcast the packets to all their own adjacent nodes (except the one they received it from) I I J L K E FH G B D C A

Janice Regan © Oct Flooding: Example  HOP 3: follow the packets I J L K E FH G B D C A

Janice Regan © Oct Flooding: Example  HOP 4: The stations receiving the packets broadcast in hop 3 broadcast the packets to all their own nearest neighbors ( not including the station they received the message from). I J L K E FH G B D C A

Janice Regan © Oct Advantages of flooding  Because packets follow every possible path, the message will get there despite link failures, so long as one path remains active (good for emergency messages)  Because packets follow every possible path at least one packet will arrive over the minimum cost route (good for establishing a virtual circuit path)  All nodes directly connected to the source will receive the message (good for getting information to all nodes)

Janice Regan © Oct Disadvantages and a simple fix  The biggest disadvantage of flooding in the volume of traffic created  If there are multiple paths to a particular node it will receive and rebroadcast the packet again and again  This creates a broadcast-storm, an increasing number of packets that continue to multiply as they travel through the network  The simplest way to prevent this is to place a short lifetime on the packet so it can only rebroadcast a few times (few time = diameter of network A better solution is sequence number controlled flooding. Each broadcast (flooded) packet is given an identifier (source id …) and a broadcast sequence number. Each node will rebroadcast a packet with a particular ID and broadcast sequence number only once

Janice Regan © Oct Sequence number controlled Flooding  HOP 1: The source station broadcasts the packet to all adjacent nodes. I I J L K E FH G B D C A

Janice Regan © Oct  HOP 1: Follow the packets I I J L K E FH G B D C A Sequence number controlled Flooding

Janice Regan © Oct  HOP 2: The receiving stations for hop 1 broadcast the packets to all their own adjacent nodes (except the one they received it from) I I J L K E FH G B D C A

Janice Regan © Oct  HOP 2: Follow the packets I J L K E FH G B D C A

Janice Regan © Oct B  HOP 3: The stations receiving the packets broadcast in hop 2 broadcast the packets to all their own nearest neighbors ( not including the station they received the message from). Nodes throw copies of the packet away and do not forward copies I J L K E FH G D C E F A

Janice Regan © Oct B  HOP 3: The stations receiving the packets broadcast in hop 2 broadcast the packets to all their own nearest neighbors ( not including the station they received the message from). Nodes throw copies of the packet away and do not forward copies I J L K E FH G D C E F A

Janice Regan © Oct  HOP 3: Follow the packets I J L K E FH G B D C B E F E F A

Janice Regan © Oct  HOP 4: The stations receiving the packets broadcast in hop 3 broadcast the packets to all their own nearest neighbors ( not including the station they received the message from). Nodes throw copies of the packet away and do not forward copies I J L K E FH G B D C B E F E F A

Janice Regan © Oct OSPF Flooding protocol  A message(LSA) contains a database record. A database record contains information about one link between two routers in the graph discussed earlier. (one link is in one direction)  Each message contains a time stamp or message number  These message numbers are used by the receiving node to determine age of the record  Send means transmit through all attached interfaces except the one on which the incoming message arrived

Janice Regan © Oct OSPF Flooding protocol  Receive message: Find the corresponding record in the local database if it exists  If the record is not yet in the local database add the record. Send the message  If the record’s message number is larger than the message number in the data base, replace the message in the database with the new record. Send the message.  If the records message number is the same as the message number in the database do nothing  If the records message number is smaller than the message number in the database, send the record in the database back through the interface on which the message arrived