[ Pobierz całość w formacie PDF ]
.Demonstrate a topology in which the spanning tree can reconfigure without someparticular bridge's being aware that the topology has changed (assuming the topologychange notification protocol was not implemented).Demonstrate how portions of thatbridge's station learning cache can become invalid as a result of that reconfiguration.8.Assume that a bridge has seven ports and that there are three protocols usingmulticast addresses MULT1, MULT2, and MULT3, respectively.Suppose you want toallow MULT1 to travel freely between all ports; you want MULT2 to travel betweenports 1, 2, and 3 and also between 5 and 7; and you want MULT3 not to travelbetween any of the ports.What network management commands would you give tothe bridge to set up the filtering?9.Where would filtering based on source address be useful? For example, a bridgemight be configured with a set of source addresses and refuse to forward a packetunless the source was in the configured set.Also, a bridge might be configured with aset of disallowed source addresses in which case, it would refuse to forward apacket whose source address was in the disallowed set.10.Imagine a protocol in which clients and servers find one another based on multicastadvertisements or solicitations.Suppose that for that protocol, there are specificregions of the extended LAN, and the person responsible for the extended LAN wantsonly clients and servers from the same region talking to one another.Assumingnonmalicious nodes, suppose that the manager configures bridges outside each ofthe regions to drop the multicast addresses used by the protocol.Furthermore,assume that the bridges do not filter based on protocol type.Will filtering based solelyon the multicast addresses keep the separate regions segregated with respect to thatprotocol?11.Discuss methods (both hardware and software) for organizing the station learningcache to speed the forwarding of packets.Adding entries to the cache based onnewly learned source addresses should also be reasonably efficient but is not ascritical as lookups of addresses already in the cache.The latter must be done at LANspeeds, whereas adding entries can be done in the background.12.Assume for simplicity that there are only two types of LANs in the topology: FDDI and802.3.FDDI allows 4,500-octet packets; 802.3 has a maximum packet size of 1,500octets.FDDI has a priority field, which is not present in 802.3.The bridge spec states thatwhen a bridge transmits a packet from 802.3 to FDDI, the bridge should transmit it atpriority = 0.It is recommended that FDDI stations not originate traffic with priority 0. Use this information to devise a simple scheme that two FDDI stations can use todiscover whether they can communicate with 4,500-octet packets or 1,500-octetpackets.(Note that they must communicate with 1,500-octet packets if the pathbetween them includes an 802.3 LAN.)13.As stated in Section 3.4.2, the original spanning tree algorithm had only a singleintermediate state between blocking and forwarding and did not indicate whether abridge should perform learning of station addresses on a port in the intermediatestate.The 802 bridge spec subdivides the intermediate state into two states.In thefirst, learning is not done; in the second, learning is done.Compare the followingthree strategies:a.A single intermediate state, in which learning is doneb.A single intermediate state, in which learning is not donec.As in the spec, two intermediate statesConsider the implications, such as unnecessarily forwarded frames and incorrectcache entries, of each strategy.14.Compare the functionality of n-port bridges with two-port bridges.Can everyconfiguration built with n-port bridges be built with two-port bridges? Suppose atopology consists of n LANs interconnected with a single n-port bridge; in thattopology, a packet always requires just a single hop to get from any LAN to any otherLAN.Can the same thing be accomplished with two-port bridges? If a customerwanted n LANs interconnected and only two-port bridges were available, whatconfiguration of LANs and two-port bridges would you suggest?15.Recently there was a proposal for a modification to the spanning tree algorithm to getrid of preforwarding delay in certain cases.The modification is that if the spanningtree algorithm tells you that link L should now be your in-link (path to the root), youare allowed to immediately start forwarding on L provided that you immediately stopforwarding on your previous in-link.The claim is that this modification will notintroduce loops.a.Can this modification ever introduce loops (even if multiple events occur invarious portions of the network and different root candidates appear)? (Hint: Ithink the answer is no.)b.Show that this proposal would introduce loops if the Designated Bridge addsthe cost of the link into its Hello message that is, the cost advertised is thecost to the LAN rather than the cost to the bridge.In other words, if B thinks itis 5 from the Root R and that link L has cost 2, B will advertise (R, 7, B, & ) inits Hello.(Isn't it amazing that such a seemingly arbitrary choice of who addsin the link cost could make a difference here?)c.Show a case in which this change causes a temporary partition when, withthe original specification, no such delay was necessary.(Hint: Themodification calls for instantaneously turning off the old in-link, whereas theoriginal said you can keep a link forwarding if it was forwarding in the oldtopology.)d.In what topologies does this modification speed forwarding of packets?Chapter 4 [ Pobierz całość w formacie PDF ]
zanotowane.pl doc.pisz.pl pdf.pisz.pl odbijak.htw.pl
.Demonstrate a topology in which the spanning tree can reconfigure without someparticular bridge's being aware that the topology has changed (assuming the topologychange notification protocol was not implemented).Demonstrate how portions of thatbridge's station learning cache can become invalid as a result of that reconfiguration.8.Assume that a bridge has seven ports and that there are three protocols usingmulticast addresses MULT1, MULT2, and MULT3, respectively.Suppose you want toallow MULT1 to travel freely between all ports; you want MULT2 to travel betweenports 1, 2, and 3 and also between 5 and 7; and you want MULT3 not to travelbetween any of the ports.What network management commands would you give tothe bridge to set up the filtering?9.Where would filtering based on source address be useful? For example, a bridgemight be configured with a set of source addresses and refuse to forward a packetunless the source was in the configured set.Also, a bridge might be configured with aset of disallowed source addresses in which case, it would refuse to forward apacket whose source address was in the disallowed set.10.Imagine a protocol in which clients and servers find one another based on multicastadvertisements or solicitations.Suppose that for that protocol, there are specificregions of the extended LAN, and the person responsible for the extended LAN wantsonly clients and servers from the same region talking to one another.Assumingnonmalicious nodes, suppose that the manager configures bridges outside each ofthe regions to drop the multicast addresses used by the protocol.Furthermore,assume that the bridges do not filter based on protocol type.Will filtering based solelyon the multicast addresses keep the separate regions segregated with respect to thatprotocol?11.Discuss methods (both hardware and software) for organizing the station learningcache to speed the forwarding of packets.Adding entries to the cache based onnewly learned source addresses should also be reasonably efficient but is not ascritical as lookups of addresses already in the cache.The latter must be done at LANspeeds, whereas adding entries can be done in the background.12.Assume for simplicity that there are only two types of LANs in the topology: FDDI and802.3.FDDI allows 4,500-octet packets; 802.3 has a maximum packet size of 1,500octets.FDDI has a priority field, which is not present in 802.3.The bridge spec states thatwhen a bridge transmits a packet from 802.3 to FDDI, the bridge should transmit it atpriority = 0.It is recommended that FDDI stations not originate traffic with priority 0. Use this information to devise a simple scheme that two FDDI stations can use todiscover whether they can communicate with 4,500-octet packets or 1,500-octetpackets.(Note that they must communicate with 1,500-octet packets if the pathbetween them includes an 802.3 LAN.)13.As stated in Section 3.4.2, the original spanning tree algorithm had only a singleintermediate state between blocking and forwarding and did not indicate whether abridge should perform learning of station addresses on a port in the intermediatestate.The 802 bridge spec subdivides the intermediate state into two states.In thefirst, learning is not done; in the second, learning is done.Compare the followingthree strategies:a.A single intermediate state, in which learning is doneb.A single intermediate state, in which learning is not donec.As in the spec, two intermediate statesConsider the implications, such as unnecessarily forwarded frames and incorrectcache entries, of each strategy.14.Compare the functionality of n-port bridges with two-port bridges.Can everyconfiguration built with n-port bridges be built with two-port bridges? Suppose atopology consists of n LANs interconnected with a single n-port bridge; in thattopology, a packet always requires just a single hop to get from any LAN to any otherLAN.Can the same thing be accomplished with two-port bridges? If a customerwanted n LANs interconnected and only two-port bridges were available, whatconfiguration of LANs and two-port bridges would you suggest?15.Recently there was a proposal for a modification to the spanning tree algorithm to getrid of preforwarding delay in certain cases.The modification is that if the spanningtree algorithm tells you that link L should now be your in-link (path to the root), youare allowed to immediately start forwarding on L provided that you immediately stopforwarding on your previous in-link.The claim is that this modification will notintroduce loops.a.Can this modification ever introduce loops (even if multiple events occur invarious portions of the network and different root candidates appear)? (Hint: Ithink the answer is no.)b.Show that this proposal would introduce loops if the Designated Bridge addsthe cost of the link into its Hello message that is, the cost advertised is thecost to the LAN rather than the cost to the bridge.In other words, if B thinks itis 5 from the Root R and that link L has cost 2, B will advertise (R, 7, B, & ) inits Hello.(Isn't it amazing that such a seemingly arbitrary choice of who addsin the link cost could make a difference here?)c.Show a case in which this change causes a temporary partition when, withthe original specification, no such delay was necessary.(Hint: Themodification calls for instantaneously turning off the old in-link, whereas theoriginal said you can keep a link forwarding if it was forwarding in the oldtopology.)d.In what topologies does this modification speed forwarding of packets?Chapter 4 [ Pobierz całość w formacie PDF ]