Routers & Routing : Computer Networking. Binary Search on Ranges. Speeding up Prefix Match - Alternatives

Transcription

1 Routers & Routing -44: omputer Networking High-speed router architecture Intro to routing protocols ssigned reading [McK9] Fast Switched ackplane for a Gigabit Switched Router Know RIP/OSPF L-4 Intra-omain Routing Speeding up Prefix Match - lternatives Route caches Temporal locality Many packets to same destination Other algorithms Waldvogel Sigcomm 9 inary search on prefixes Works well for larger addresses remler-arr Sigcomm 99 lue = prefix length matched at previous hop Why is this useful? inary Search on Ranges Prefixes P = *, P = *, P = * P P P P P P - > - = - P P P P ncode each prefix as range and place all range endpoints in binary search table or tree. Need two next hops per entry for > and = case. [Lampson, Srinivasan, Varghese] Problem: Slow search (log N+ = for a million prefixes) and update (O(n)). 4

2 Speeding up Prefix Match - lternatives ontent addressable memory (M) Hardware based route lookup Input = tag, output = value associated with tag Requires exact match with tag Multiple cycles ( per prefix searched) with single M Multiple Ms ( per prefix) searched in parallel Ternary M,,don t care values in tag match Priority (I.e. longest prefix) by order of entries in M Outline Packet classification IP router design Routing protocols distance vector Routing protocols link state 6 Packet lassification Typical uses Identify flows for QoS Firewall filtering Requirements Match on multiple fields Strict priority among rules.g. no traffic from..*. ok traffic on port omplexity N rules and k header fields for k > O(log N k- ) time and O(N) space O(log N) time and O(N k ) space How many rules? Largest for firewalls & similar iffserv/qos much larger k (?)

3 it Vectors it Vectors Rule Field * Field * Rule Field * Field * * * * * * * * * * * Field * * Field 9 Observations [GM99] ommon rule sets have important/useful characteristics Packets rarely match more than a few rules (rule intersection).g., max of 4 rules seen on common databases up to rules Special cases for k = source and destination O(log N) time and O(N) space solutions exist ggregating Rules [V] ommon case: very few s in bit vector aggregate bits OR together bits at a time N/ bit-long vector typically chosen to match word-size an be done hierarchically aggregate the aggregates N of aggregate bits indicates which groups of rules have a possible match Hopefully only a few s in N ed vector N of aggregated bit vectors may have false positives Fetch and N just bit vectors associated with positive entries

4 Rearranging Rules [V] Problem: false positives may be common Solution: reorder rules to minimize false positives What about the priority order of rules? How to rearrange? Heuristic sort rules based on single field s values First sort by prefix length then by value Moves similar rules close together reduces false positives Summary: ddressing/lassification Global addressing matches design objectives of Internet Speed of forwarding lookup/classification of packets is a key challenge for routers specially high-speed backbone routers IR provided more structured routing Good examples of using common case optimization Routing with a clue lassification with few matching rules 4 Outline What oes a Router Look Like? Packet classification IP router design Routing protocols distance vector Network controller Line cards ackplane Routing protocols link state 6 4

5 Network Processor Runs routing protocol and downloads forwarding table to line cards Some line cards maintain two forwarding tables to allow easy switchover Performs slow path processing Handles IMP error messages Handles IP option processing Line ards Network interface cards Provides parallel processing of packets Fast path per-packet processing Forwarding lookup (hardware/si vs. software) Switch esign Issues Have N inputs and M outputs Multiple packets for same output output contention Switch contention switch cannot support arbitrary set of transfers rossbar us High clock/transfer rate needed for bus anyan net omplex scheduling needed to avoid switch contention Solution buffer packets where needed Switch uffering Input buffering Which inputs are processed each slot schedule? Head of line packets destined for busy output blocks other packets Output buffering Output may receive multiple packets per slot Need speedup proportional to # inputs Internal buffering Head of line blocking mount of buffering needed 9

6 Line ard Interconnect ISLIP Virtual output buffering Maintain per output buffer at input Solves head of line blocking problem ach of MxN input buffer places bid for output rossbar connect hallenge: map of bids to schedule for crossbar ISLIP (cont.) Summary - IP router design ifferent architectures for different types of routers High speed routers incorporate large number of processors ommon case is optimized carefully 4 6

7 Outline Factors ffecting Routing Packet classification IP router design Routing protocols distance vector Routing algorithms view the network as a graph Problem: find lowest cost path between two nodes Factors Static topology ynamic load Policy F Routing protocols link state 6 Two Main pproaches istance-vector (V) protocols Link state (LS) protocols istance Vector Protocols mployed in the early rpanet istributed next hop computation Unit of information exchange Vector of distances to destinations istributed ellman-ford lgorithm

8 The ouncing ffect Sends Routes to dest cost X dest cost 9 Updates istance to Sends Routes to dest cost dest cost 4

9 Sends Routes to How are These Loops aused? dest cost Observation : s metric increases Observation : picks as next hop to ut, the implicit path from to includes itself! 4 4 Solution : Holddowns If metric increases, delay propagating information In our example, delays advertising route eventually thinks s route is gone, picks its own route then selects as next hop dversely affects convergence Other Solutions Split horizon does not advertise route to Poisoned reverse advertises route to with infinite distance Works for two node loops oes not work for loops with more nodes 6 9

10 xample Where Split Horizon Fails voiding the ouncing ffect X When link breaks, marks as unreachable and reports that to and Suppose learns it first now thinks best path to is through reports unreachable to and a route of cost= to thinks is reachable through at cost 4 and reports that to reports a cost to who reports new cost to etc... Select loop-free paths One way of doing this: ach route advertisement carries entire path If a router sees itself in path, it rejects the route GP does it this way Space proportional to diameter Loop Freedom at very Instant oes bouncing effect avoid loops? No! Transient loops are still possible Why? ecause implicit path information may be stale See this in GP convergence Only way to fix this nsure that you have up-to-date information by explicitly querying istance Vector in Practice RIP and RIP Uses split-horizon/poison reverse GP Propagates entire path Path also used for effecting policies 9 4

11 Outline Packet classification IP router design Routing protocols distance vector asic Steps Start condition ach node assumed to know state of links to its neighbors Step ach node broadcasts its state to all other nodes Reliable flooding mechanism Step ach node locally computes shortest paths to all other nodes from global state ijkstra s shortest path tree (SPT) algorithm Routing protocols link state 4 4 Link State Packets (LSPs) Periodically, each node creates a link state packet containing: Node I List of neighbors and link cost Sequence number Needed to avoid stale information from flood Time to live (TTL) Node outputs LSP on all its links Reliable Flooding When node J receives LSP from node K If LSP is the most recent LSP from K that J has seen so far, J saves it in database and forwards a copy on all links except link LSP was received on Otherwise, discard LSP How to tell more recent Use sequence numbers Same method as sliding window protocols 4 44

12 Link State haracteristics With consistent LSs, all nodes compute consistent loop-free paths Limited by ijkstra computation overhead, space requirements an still have transient loops X Packet from may loop around if knows about failure and & do not Summary: LS vs. V In V send everything you know to your neighbors In LS send info about your neighbors to everyone Msg size: small with LS, potentially large with V Msg exchange: LS: O(n), V: only to neighbors 4 46 Summary: LS vs. V onvergence speed: LS: faster don t need to process LSPs before forwarding V: fast with triggered updates Space requirements: LS maintains entire topology V maintains only neighbor state Summary: LS vs. V Robustness: LS can broadcast incorrect/corrupted LSP an be made robust since sources are aware of alternate paths V can advertise incorrect paths to all destinations Incorrect calculation can spread to entire network 4 4

13 Summary: LS vs. V In LS nodes must compute consistent routes independently - must protect against LS corruption In V routes are computed relative to other nodes ottom line: no clear winner, but we see more frequent use of LS in the Internet Next Lecture: Inter-omain Routing order Gateway Protocol (GP) ssigned reading [L] elayed Internet Routing onvergence [Nor] Internet Service Providers and Peering 49 istributed ellman-ford xample - Initial istances Start onditions: ach router starts with a vector of (zero) distances to all directly attached networks Send step: ach router advertises its current vector to all neighboring routers Info at Node istance to Node Receive step: Upon receiving vectors from each of its neighbors, router computes its own distance to each neighbor Then, for every network X, router finds that neighbor who is closer to X than any other neighbor Router updates its cost to X fter doing this for all X, router goes to send step

14 Receives s Routes; Updates ost receives s; Updates ost Info at Node istance to Node Info at Node istance to Node receives s routes; Updates osts Final istances Info at Node istance to Node Info at Node istance to Node

15 View From a Node Final istances fter Link Failure s routing table Next hop Info at Node istance to Node dest X 9 9 SPT lgorithm (ijkstra) SPT xample SPT = {a} for all nodes v if v adjacent to a then (v) = cost (a, v) else (v) = infinity Loop find w not in SPT, where (w) is min add w in SPT for all v adjacent to w and not in SPT (v) = min ((v), (w) + (w, v)) until all nodes are in SPT F step SPT (b), P(b) (c), P(c) (d), P(d) (e), P(e) (f), P(f),,, F 9 6

16 xample xample F step SPT (b), P(b) (c), P(c) (d), P(d) (e), P(e) (f), P(f),,,, 4,, F F step SPT (b), P(b) (c), P(c) (d), P(d) (e), P(e) (f), P(f),,,, 4,,,, 4, F 6 6 xample xample F step SPT (b), P(b) (c), P(c) (d), P(d) (e), P(e) (f), P(f),,,, 4,,,, 4,, 4, F F step SPT (b), P(b) (c), P(c) (d), P(d) (e), P(e) (f), P(f),,,, 4,,,, 4,, 4, 4 4, F

17 xample F step SPT (b), P(b) (c), P(c) (d), P(d) (e), P(e) (f), P(f),,,, 4,,,, 4,, 4, 4 4, F F 6