Examination IP routning inom enkla datornät, DD2490 IP routing in simple networks, DD2490 KTH/CSC Date: 20 May 2009 14:00 19:00 SOLUTIONS a) No help material is allowed - You are not allowed to use books or calculators! b) You may answer questions in English or in Swedish. c) Put a mark in the table on the cover page for each question you have answered. d) Max points is 50, Limit for passing (Grade E) is 25 e) Course responsible is Olof Hagsand, phone 08-790 6534
1. Routers and routing (10p) Answer briefly: 1. With respect to data traffic in a router: What is slow-path? What is fast path? (1p) Packets are forwarded on the slow path if they need special processing so that they need to pass through the route processor. Traffic forwarded using the regular forwarding path directly from line-card to line-card takes the fast path. 2. What is Ternary Content-Addressable Memory (TCAM) and how can it be used in a router? (1p) TCAMs are examples of memories where the lookup is based on the contents instead of the address. In a router the content can be an IP address and don't care values are used to represent the network mask. Thus, IP lookup can be made using TCAMs in a single cycle. 3. Address lookup in IPv4 is made doing longest prefix match. But what is meant with packet classification and name one major usage? (1p) Packet classification is a more general method where more fields in the packet header are used to sort packets into different classes. Major usages are in packet filtering, firewalls, traffic engineering, quality of service handling: metering, shaping, etc. 4. Routers may be partitioned into virtual (or logical) routers. Name one usage of virtual routers in networking? (1p) In applications where there is a need for separate address spaces. This includes different VPN applications including L3VPN, pseduowire, VPLS, etc. 5. What is little-endian byte-ordering and in what way may it be of importance when writing forwarding code? (1p) Little endian is when a CPU handles integers using the least significant byte first (example is Intel i386 CPUs used in PCs). A protocol implementation that uses binary fields of more than two bytes needs to handle this by byte-swapping since most protocols use big endian byte order as an inter-operability format. 6. What is the hot-potato routing policy and explain why it often leads to asymmetric routing? (1p) Hot-potato is a routing policy where traffic leaves a routing domain as fast as possible. This is the default BGP policy and happens if external routes are preferred over internal routes. If two adjacent routing domains with more than one inter-connection both use hotpotato, there is a high probability that this will cause asymmetric routing. Since both will try send their traffic on the first connection that leaves the domain.
7. Static/aggregated routes are often configured without next-hops, or outgoing interfaces, so that there is no explicit way to forward traffic. What is the purpose of such routes? (1p) They attract traffic, the traffic is dropped if there are no more specific routes to follow. Therefore there is no need to specify nexthop, instead a drop/reject/discard rule is associated with the route. 8. Why does a direct route have lower precedence than a route from a dynamic routing protocol? (1p) If a destination is reachable on a directly connected network, there can be no other route learnt from a routing protocol that can be better. 9. What is RPF (Reverse Path forwarding) method and when is it used? (1p) In RPF, the source of a packet is looked up in a routing table. It is used in multicast forwarding as a simple flooding rule: If a packet is received on the same interface that would have been used for forwarding if the address was a destination, then it is forwarded on other interfaces. Other usages are in address spoofing avoidance: only admit packets from an interface whose source addresses are reachable on that interface. 10. What is the purpose of IGMP Snooping? (IGMP - Internet Group Management Protocol) (1p) To avoid flooding multicast traffic on a bridge network. Listen to IGMP messages and thereby prune multicast delivery only to listening receivers.
2. Addressing (5p) Regard the network in the figure in Appendix A consisting of routers R1- R14 interconnected by Ethernet links with interface names (,..) as shown. There is an upper network connecting routers R1,R2,...,R7, a backbone network connecting R1 and R8, and a lower network connecting R8,R9,...,R14. Your task is to assign an IP-subnet to each network, and an IP address to each interface in the figure. Note the following: All Ethernet links connect two routers, except the two networks connecting R1, R2, R3 and R8, R9, R10, respectively. They are shared networks (one Ethernet broadcast domain) with one IP subnet each connecting three routers. Use /30 networks for the point-to-point links, and /29 for the links connecting three routers. The router-ids are given by the figure: The router-id of R1 is 1, router-id of R2 is 2, etc. The address assignment of the backbone link between R1 and R8 is given in the figure as 10.0.0.64/30. Likewise, 10.0.0.0/30 is given to the link connecting R2 and R4. Interface addresses are given as shown in the figure. You should allocate addresses so that it is possible to aggregate all sub-ip networks of the upper network into 10.0.0.0/27 and to aggregate the lower network into 10.0.0.32/27. There may be holes in the address block, but no address of the block may appear in another network. For example, no sub-network of 10.0.0.0/27 may appear in the lower network. All link metrics are equal to one. Draw the network with your address blocks and interface addresses (as the pre-configured addresses in the figure) which clearly shows the assignments you have made. You will then re-use this assignment in later exercises so it is important to get it right. You can use Appendix A to draw the network and submit with your exam. (5p) Example address allocation shown. There are other possibilities.
upper R2 R4 R5 R6 R7.1.5.9.13 10.0.0.0/30 10.0.0.4/30 10.0.0.8/30 10.0.0.12/30 e1.2 e2 e1.10 e2.14.6 backbone R1 10.0.0.64/30 R8.65.66 R3.17 10.0.0.16/29.18 e1.19 e1.51.49 10.0.0.48/29.50 lower R9 R10 e1.34 e2.38 e1.42 e2.46 10.0.0.32/30 10.0.0.36/30 10.0.0.40/30 10.0.0.44/30.33.37.41.45 R11 R12 R13 R14 3. Distance-Vector and RIP (5p) Assume you run RIPv2 (Routing Information Protocol) on the upper network of Appendix A (R1-R7). (You must have assigned addresses in exercise 2 first) 1. Show the routing table of all RIP routes in R1. The routing table should contain destination prefix, next-hop and metric. Note that you should show all RIP routes even though they may be overridden in the router by for example direct routes. (2p) Destination next-hop/i-f metric ----------------------------------------------------------------------- 10.0.0.0/30 10.0.0.17 2 10.0.0.4/30 10.0.0.17 2 10.0.0.8/30 10.0.0.18 2 10.0.0.12/30 10.0.0.18 2 10.0.0.16/29 10.0.0.17 2 10.0.0.16/29 10.0.0.18 2 (alt only one if no load-balancing) 2. Assume split-horizon with poison reverse is enabled on all routers. Which RIP route updates will router R3 get on interface? That is, which distance vectors (prefix, metrics) will R3 get from each router R1 and R2? (3p) From R1
destination metric -------------------------------------- 10.0.0.0/30 16 10.0.0.4/30 16 10.0.0.8/30 16 10.0.0.12/30 16 10.0.0.16/29 1 From R2: destination metric -------------------------------------- 10.0.0.0/30 1 10.0.0.4/30 1 10.0.0.8/30 16 10.0.0.12/30 16 10.0.0.16/29 1 4. OSPF (5p) Assume you run OSPFv2 (Open Shortest Path First), on the lower network in the network in Appendix A. You must have assigned addresses in exercise 2 first. Assume that R1 also runs OSPF and redistributes the RIP routes from exercise 3 as external routes using E1 (external type 1). All routers are in area 0. Show the OSPF LSAs as they appear in router R12 after the protocol has stabilized. Specify for all LSAs: LS type, LS identifier, and advertizing router. For router LSA, specify link type and link id for each associated link. For network LSAs specify netmask and connected router ids. For summary and external LSAs, specify mask and metric. (5p) NOTE 1: Some have modelled the Ethernet with two end-points as OSPF p-t-p, others have modelled them as transit. Both are correct in principle. NOTE 2: The RIP routes are either aggregated to one /27 with metric 2 (max of all sub-routes) or the five individual routes. The two alternatives follow Alternative 1 using router LSA transit links LS type LS id Adv router link type, link id ----------------------------------------------------------------------------- 1 1 1 transit 10.0.0.65 (If R1 is DR) 1 8 8 transit 10.0.0.65 transit 10.0.0.49 (If R9 is DR) 1 9 9 transit 10.0.0.49 transit 10.0.0.33 (If R11 is DR) transit 10.0.0.37 (If R12 is DR) 1 10 10 transit 10.0.0.49 transit 10.0.0.41 (If R13 is DR) transit 10.0.0.45 (If R14 is DR) 1 11 11 transit 10.0.0.33 1 12 12 transit 10.0.0.37 1 13 13 transit 10.0.0.41
1 14 14 transit 10.0.0.45 2 10.0.0.65 1 1, 8 netmask: /30 2 10.0.0.49 8 8, 9, 10 netmask: /29 2 10.0.0.33 9 9, 11 netmask: /30 2 10.0.0.37 9 9, 12 netmask: /30 2 10.0.0.41 10 10, 13 netmask: /30 2 10.0.0.45 10 10, 14 netmask: /30 5 10.0.0.0 1 netmask: /27 metric: 2, type: E1 Alternative 2 using p-t-p router LSA links: LS type LS id Adv router link type, link id ----------------------------------------------------------------------------- 1 1 1 p-t-p 8 (points to R8) stub 10.0.0.66, (alt 10.0.0.64/30) 1 8 8 p-t-p 1 stub 10.0.0.65, (alt 10.0.0.64/30) transit 10.0.0.49 (If R9 is DR) 1 9 9 transit 10.0.0.49 p-t-p 11 stub 10.0.0.33, (alt 10.0.0.32/30) p-t-p 12 stub 10.0.0.37, (alt 10.0.0.34/30) 1 10 10 transit 10.0.0.49 p-t-p 13 stub 10.0.0.41, (alt 10.0.0.40/30) p-t-p 14 stub 10.0.0.45, (alt 10.0.0.44/30) 1 11 11 p-t-p 9 stub 10.0.0.34, (alt 10.0.0.32/30) 1 12 12 p-t-p 9 stub 10.0.0.38, (alt 10.0.0.36/30) 1 13 13 p-t-p 10 stub 10.0.0.42, (alt 10.0.0.40/30) 1 14 14 p-t-p 10 stub 10.0.0.46, (alt 10.0.0.44/30) 2 10.0.0.48 8 8, 9, 10 netmask: /29 5 10.0.0.0 1 netmask: /27 metric: 2, type: E1 Alternative external routes (for both above) 5 10.0.0.0 1 netmask: /30 metric: 2, type: E1 5 10.0.0.4 1 netmask: /30 metric: 2, type: E1 5 10.0.0.8 1 netmask: /30 metric: 2, type: E1 5 10.0.0.12 1 netmask: /30 metric: 2, type: E1 5 10.0.0.16 1 netmask: /29 metric: 1, type: E1
5. OSPF areas (5p) Assume that you run OSPF on the whole network in Appendix A (No RIP). You must have assigned addresses in exercise 2 first. Assign area 1 to the upper network and area 2 to the lower network. The backbone network is area 0. R1 is ABR (Area Border Router) between area 0 and 1 and R8 is ABR between area 0 and area 2. Assume R1 aggregates the upper network and announces 10.0.0.0/27 from area 1 and R8 aggregates and announces 10.0.0.32/27 from area 2. Show the LSAs in router R12. (Same information as specified in exercise 4)(5p) LS type LS id Adv router link type, link id ----------------------------------------------------------------------------- 1 8 8 transit 10.0.0.49 (Same LSAs type 1 for routers 9-14 as in exercise 4) (Same LSAs tif alternative 1: same type 2 except 10.0.0.65) 3 10.0.0.0 8 netmask: /27 3 10.0.0.64 8 netmask: /30 (No type 5 LSAs) 6. More link-state routing (6p) 1. Suppose an OSPF routers stops working instantly without a chance to communicate with is neighbours. How do the other (neighboring) routers know that it is dead? (Name at least two methods) (2p) Either by (1) OSPF hello messages sent every Hello interval (default 10s. After a number of lost hellos, Router dead Interval, the neighbor is considered dead. (2) Link detection - if the link is directly connected between the routers, the neighboring router may immediately detect that the link fails, and thus that the neighboring roure is not reachable (3) BFD (Bidirectional Forwarding Meschanisms). (4) There are also other, vendor-specific methods. 2. The LSAs that the dead router has announced will eventually be deleted by the routers in the network. How does this mechanism work, how long does it take (approximately), and which fields in the OSPF LSA header controls this? (2p) The LS age field in the LSA header controls the lifetime of the LSAs. In worst case, the LSA will continue to live to MaxAge which is default 60 minutes. Under normal circumstances, the originator router would increment the sequence number and re-introduce the LSA with the incremented sequence number after half this time. Assume the network in exercise 5 used IS-IS (Intermediate-System to Intermediate-System) instead of OSPF as routing protocol. What differences would this lead to with respect to areas? Describe how an area configuration of exercise 5 could be configured using IS-IS. (2p)
There can be several solutions. If area limit is between R1 and R8, then: R1 L1/L2 areaid 1 R8 L1/L2 areaid 2 R2-R7 L1 areaid 1 R9-R14 L1 areaid 2 In this case, R1 and R8 can aggregate the L1 routes when distributing them into L2. The L2 routes will then be redistributed into L1. Alternatively, the interior routers can all be L1/L2 R1 L2 R8 L2 R2-R3 L1/L2 areaid 1 R9-R10 L1 /L2 areaid 2 R4-R7 L1 areaid 1 R11-R14 areaid 2 With R2-R3 and R9-R10 doing L1/L2 redistribution. 7. Multicast (6p) Assume you have the network of Appendix A running OSPF or IS-IS as an IGP (Interior Gateway Protocol) and you start PIM-SM (Protocol Independent Multicast - Sparse Mode) on all routers R1-R14. You allocate two RP:s, R1 and R8, and use anycast-rp, where the anycast address used for this purpose is, say, 10.0.0.68. The RP:s communicate using MSDP (Multicast Source Discovery Protocol). Both RP:s serve all multicast groups. 1. First, router R11 receives an IGMP membership report on one of its interfaces indicating that a host has made a (*,241.2.3.4) join. This will result in PIM-SM state being established in the network. Which PIM-SM messages are sent and show the multicast forwarding state in all affected routers using a table with entries: Router, (S,G), oiflist (output interface list) (2p) Note: 241.2.3.4 is an illegal multicast address. This was an exam mistake. The answer is based on the assumption that the address is legal. The student must otherwise have pointed this out. Router (S,G) oiflist ----------------------------------------------------------------- 8 (*,241.2.3.4) e1 9 (*,241.2.3.4) e1 11 (*,241.2.3.4) x (the interface is unknown) 2. Then, R7 starts receiving multicast traffic destined to 241.2.3.4 from sender 10.1.1.1 on one of its interfaces. Describe what kind of trees will be created, which PIM-SM messages are sent, and which MSDP messages are sent as a consequence of this? (2p) A PIM-register message is sent from the DR (R7) to the RP(R1). An SPT tree is built from R7 to RP after the RP has sent a PIM-Join (10.1.1.1,241.2.3.4) towards the DR. An MSDP SA message will be
sent from R1 to R8, so that R8 in turn will build an SPT frm R8 to R7. Finally, when the first multicast data messages reach R11, R11 will send a PIM-Join (10.1.1.1,241.2.3.4) towards R7, and an SPT will be built from R7 to R11. The (S,G) tree will in principle be pruned from R1, when this final tree is constructed, but since it takes the same path, the state will remain. 3. Show the multicast forwarding state in all affected routers in the network after the final trees have been constructed. That is, after eventual tree switchover. Use the same table format as in the exercise above. (2p) Router (S,G) oiflist ----------------------------------------------------------------- 7 (10.1.1.1, 241.2.3.4) 3 (10.1.1.1, 241.2.3.4) 1 (10.1.1.1, 241.2.3.4) 8 (*,241.2.3.4) e1 (10.1.1.1, 241.2.3.4) e1 9 (*,241.2.3.4) e1 (10.1.1.1, 241.2.3.4) e1 11 (*,241.2.3.4) x (the interface is unknown) (10.1.1.1, 241.2.3.4) x
8. Spanning-Tree (3p) 4 1 3 2 1 2 2 3 3 1 3 1 2 3 1 2 Study the switched network with learning bridges running spanning-tree above with bridge IDs and port numbers as indicated. All link metrics have the value one. Note the shared network between 1 and 3 which is a shared ethernet with four ports connected to it. Show (or describe) the spanning-tree of the network, indicating root, designated and blocked ports. (3p) 9. MPLS (5p) Describe how traffic engineering using RSVP/MPLS (Resource Reservation protocol/ Multi-protocol label switching) for transit traffic in a BGP environment works. You should cover: Steering of traffic paths. Resource reservation. Encapsulation of transit traffic in MPLS The role of IGP vs MPLS vs RSVP vs BGP How transit traffic is sent (selected) via MPLS LSPs What advantages using MPLS for transit over IBGP. The consequences of using next-hop self or not in BGP. When MPLS Labelled Switch Paths(LSPs) are used for transit traffic, it is possible to control which paths the LSPs take by using RSVP source routing. By source routing it is possible to steer the LSPs via certain routers (loose source routing), or even to pin the path explicitly (strict source routing). Using the reservation mechanisms in RSVP, in particularthe bandwidth reservation, it is possible to reserve network resources for these paths. Traffic sent in such LSPs are encapsulated using a single MPLS header (single stack) with the labels determined by RSVP during the LSP signaling. Typically, the routers switch labels from the ingress to the egress while the pen-ultimate router pops the label
using implicit-null assignment. In this way, the egress router does a regular IP lookup. The roles of the protocols are as follows. The IGP (eg IS-IS or OSPF) computes all shortest path routes between all internal destinations in the network. MPLS is used as a carrier of the transit data traffic. RSVP is used to setup the MPLS LSPs thorugh the internal network using the IGP to find the paths. BGP finally is used for external routes and to bind the external (transit) traffic to the MPLS LSPs established. Since only traffic matching BGP routes will be sent via MPLS LSPs, at the ingress router, it will only be external traffic entering the network destined to outside the network that will be sent via MPLS - and this is exactly the transit traffic. The advantage of using MPLS over IBGP for handling transit traffic is that no BGP speakers need to exist in the interior of the network. Thus, the major drawback of IBGP, full mesh peering, is avoided, as well as the extra cost of having BGP capable internal routers. It is also easier to steer transit traffic using source routing. In order for transit traffic to be bound to LSPs it is necessary for the BGP routes to have the same next-hop as the remote LSP end-points. But LSPs may only be setup between border routers (MPLS is almost always used within a network), and the only way to ensure that BGP nexthop is the egress router is to use next-hop self. If BGP next-hop self is not used, the BGP next-hop is an external address (the address of the external peer) and the MPLS LSP may not terminate there.
Appendix A: Network topology for exercises 2-7 Note that you should allocate address blocks in exercise 2 before using the network in the other exercises. R4 R5 R6 R7.1 10.0.0.0/30 e1.2 e2 e1 e2 upper R2 R3 backbone lower e1 R1 10.0.0.64/30 R8.65.66 e1 R9 R10 e1 e2 e1 e2 R11 R12 R13 R14