1. INF5050: Protocols and routing in the internet By: Desta Haileselassie Hagos February 13, 2015 2002

2. Recap…IP Packets Routing Switching –Two technologies Circuit Switching Packet Switching INF5050, University Of Oslo2 Network (IP) Application Transport Link Physical IP Packets

3. Scientific Contribution Issues of building next generation routers Presents several algorithms/architectures to implement IP route lookup, packet classification, and switch fabrics Surveys various switch architectures of commercial routers and switch chip sets Outlines several challenging issues that remain to be researched for next generation routers INF5050, University Of Oslo3

4. New proposal from the paper? It doesn’t propose any new system –Instead, it surveys an over all picture of next generation routers INF5050, University Of Oslo 4

5. Agenda Background: Introduction Packet Lookup and Classification Switching Fabrics INF5050, University Of Oslo5

6. Introduction (1/4) Need for next generation routers –To accommodate the bandwidth demands –Support QoS –Reduce costs Keep pace with the Internet traffic growth – Exploring transmission and switching technologies Optical transmission technologies –DWDM (Dense Wave Division Multiplexing) –Optical add-drop multiplexers –Ultralong-haul lasers –Optical amplifiers »Large impact in lowering costs of digital transmission INF5050, University Of Oslo6

7. Continued…Introduction (2/4) Current router technologies in the market –Cannot provide large switching capacities to satisfy current and future bandwidth demands Next-generation routers –High-speed interfaces (10 or 40 Gb/s) –Large switching capacity (e.g. Multipetabit) INF5050, University Of Oslo7

8. Continued…Introduction (3/4) Current approaches –IP routers and MPLS switches at the network edge Connected to OXCs – but connections are high bit rates(10/40) Interconnected by DWDM transmission equipment If edge routers want to communicate with all other routers –Direct connections to those routers Low link utilization –Connect through multiple logical hops (i.e., routed by other routers). This results in larger latency INF5050, University Of Oslo8

9. Continued….Introduction (4/4) Solutions –Build very large IP routers or MPLS switches at points of presence (POPs) Aggregate traffic from edge routers onto high-speed links that are then directly connected to other large routers at different POPs through DWDM transmission equipment –Advantage Achieves higher link utilization and fewer hops (lower latency) INF5050, University Of Oslo9

10. Continued… cost-effective Solution Replacing a cluster of mid-size routers with a large-capacity scalable router INF5050, University Of Oslo10

11. Packet Switch What are we going to talk about? –Designing a large capacity packet switch Design issues –IP Routers and routing models –The Evolution of IP Routers INF5050, University Of Oslo11

12. Designing a large capacity packet switch Can we cost-effectively build a very large IP router, with a capacity of a few petabit/s? Requirement for a packet switching –Process packets (such as classification, table lookup, and packet header modification) –Store packets –Schedule packets –Perform buffer management INF5050, University Of Oslo12

13. Designing issues (1/2) Demanding technologies when designing a large-capacity packet switch –Memory Bottleneck for large-capacity networks –Interconnection technology Significantly effects a system’s power consumption and cost INF5050, University Of Oslo13

14. Continued….Designing issues (2/2) Designing challenges for a switch architecture –Scalable to a very large capacity –Cost-effective Memory Speed  Moore ’ s law (* Prof. Tore’s lecture) Packet arbitration  resolving output port contention among the input ports QoS Control  providing QoS guarantees Optical interconnection  installation costs Power consumption *Flexibility (added from the book but not in the paper) INF5050, University Of Oslo14

15. IP Routers 6 7 8 5 4 3 1 2 12 10 13 11 Router  The Internet is a mesh of routers  Router consists of -Set of input interfaces where packets arrive -Set of output interfaces from which packets depart -Some form of interconnect connecting inputs to outputs  Router implements (1) Forward packet to corresponding output interface (2) Manage bandwidth and buffer space resources INF5050, University Of Oslo15

16. IP Routing Models Router architectures can be broadly classified INF5050, University Of Oslo16 Distributed Centralized

17. Generic Architecture Input and output interfaces are connected through an interconnect Interconnect can be implemented by –Shared memory Low capacity routers (e.g., PC-based routers) –Shared bus Medium capacity routers –Point-to-point (switched) bus High capacity routers input interface output interface Inter- connect INF5050, University Of Oslo17

18. Evolution of IP Routers Generation of IP Routers –1 st Generation IP Routers ( Shared Memory) –2 nd Generation IP Routers ( Shared Bus) –3 rd Generation IP Routers (Point-to-Point Switch) –4 th Generation IP Routers (Clustering/Multistage ) Remark  Read Professor Tor’s slide from Feb.06 lecture for more details INF5050, University Of Oslo18

19. Structure of A Generic Communication Switch Switches –Circuit switch –Ethernet switch –ATM switch –IP router Line Cards Switch Fabric Control Processor Line Cards Switch fabric –high capacity interconnect Line card –address lookup in the data path (forwarding) –*Components of line cards (Read page 5 of the paper) Control Processor –load the forwarding table (routing or signaling) INF5050, University Of Oslo19

20. Agenda Background: Introduction Packet Lookup and Classification Switching Fabrics INF5050, University Of Oslo20

21. IP Route Lookup What are we going to talk about? –Algorithms and techniques that an IP router uses to decide where to forward the packets next (routing lookup) –Algorithms and techniques that a flow-aware router uses to classify packets into flows (packet classification) INF5050, University Of Oslo21

22. RFC 1812: Requirements for IPv4 Routers Must perform an IP datagram forwarding decision (called forwarding) Must send the datagram out the appropriate interface (called switching) Remark: Optionally: a router MAY choose to perform special processing on incoming packets INF5050, University Of Oslo22

23. Examples of special processing Filtering packets for security reasons Delivering packets according to a pre-agreed delay guarantee Treating high priority packets preferentially Maintaining statistics on the number of packets sent by various routers INF5050, University Of Oslo23

24. Special Processing Requires Identification of Flows All packets of a flow obey a pre-defined rule and are processed similarly by the router E.g. a flow = (src-IP-address, dst-IP-address), or a flow = (dst-IP-prefix, protocol) etc. Router needs to identify the flow of every incoming packet and then perform appropriate special processing INF5050, University Of Oslo24

25. Flow-aware vs Flow-unaware Routers Flow-aware router: –Keeps track of flows and perform similar processing on packets in a flow Flow-unaware router (packet-by-packet router): –Treats each incoming packet individually INF5050, University Of Oslo25

26. Routing Lookups INF5050, University Of Oslo26

27. Routing –Finding and setting up routing tables Uses routing tables computed by routing protocols –Routing tables summarize information by maintaining a mapping between IP address prefixes and output interfaces Can be a complex procedure –Very large routing tables –Dynamic variation of routes in the Internet INF5050, University Of Oslo27

28. IP Routers Metrics for Lookup algorithms Metrics for lookup algorithms –Speed (Lookup time) = number of memory accesses –Storage space (amount of memory) –Low update time (support >10k updates/s) –Low preprocessing time –Ability to handle large routing tables –Flexibility in implementation –Scalability INF5050, University Of Oslo28

29. 29 Lookup in an IP Router IPv4 unicast destination address based lookup Dstn Addr Next Hop ---- Dstn-prefixNext Hop Forwarding Table Next Hop Computation Forwarding Engine Incoming Packet HEADERHEADER INF5050, University Of Oslo

30. Forwarding Engine header payload Packet Router Destination Address Outgoing Port Dest-network Port Forwarding Table Routing Lookup Data Structure 65.0.0.0/8 128.9.0.0/16 149.12.0.0/19 3 1 7 INF5050, University Of Oslo30

31. 31 Delivery with routing tables to: 20.2.1.2

32. Routing Tables Each router and each host keeps a routing table which tells the router how to process an outgoing packet Main columns –Destination address Where is the IP datagram going to? –Next hop interface How to send the IP datagram? Routing tables are set so that a datagram gets closer to the its destination INF5050, University Of Oslo32

33. Forwarding Packet forwarding –Decide to which output interface to forward each packet based on the information in packet header Examine packet header Lookup in forwarding table Update packet header INF5050, University Of Oslo33

34. Forwarding Decisions ATM and MPLS switches –Direct Lookup approach Bridges and Ethernet switches –Associative Lookup –Hashing –Trees and tries IP Router Lookup and Lookup Schemes –Caching –Patricia trees/tries –Other methods Packet Classification INF5050, University Of Oslo34

35. ATM and MPLS Switches Direct Lookup (Port, vci/MPLS label) Address Memory Data (Port, vci/label) INF5050, University Of Oslo35 *VCI = Virtual Channel Identifier *VPI = Virtual Path Identifier *VCC (Virtual Channel Connection) Incoming Port Outgoing Port

36. More on Lookups … Read Class-based addressing –Lookups with class-based addressing Classless addressing –CIDR –Size of Routing tables –Routing lookups with CIDR Routing lookups in hardware INF5050, University Of Oslo36

37. Forwarding Decisions ATM and MPLS switches –Direct Lookup Bridges and Ethernet switches –Associative Lookup –Hashing –Trees and tries IP Router Lookup and Lookup Schemes –Caching –Patricia trees/tries –Other methods Packet Classification INF5050, University Of Oslo37

38. Bridges and Ethernet Switches Associative Lookups INF5050, University Of Oslo Network Address Associated Data Associative Memory or CAM Search Data 48 bits log 2 N Associated Data Hit? Address { Advantages: Simple Disadvantages Slow High Power Small Expensive 38 *Associative memory (aka Content Addressable Memory, CAM) compares all entries in parallel against incoming data

39. Use a pseudo-random hash function (relatively insensitive to actual function) Bucket linearly searched (or could be binary search, etc.) Leads to unpredictable number of memory references Hashing Function Memory Address Data 48 16 Pointer Memory Address Data List/Bucket List of network addresses in this bucket Search Data Bridges and Ethernet Switches Hashing log 2 N Associated Data INF5050, University Of Oslo39

40. Forwarding Decisions ATM and MPLS switches –Direct Lookup Bridges and Ethernet switches –Associative Lookup –Hashing –Trees and tries IP Route Lookup and Lookup Schemes –Caching –Patricia trees/tries –Other methods Packet Classification INF5050, University Of Oslo40

41. Lookup Identify the output interface to forward an incoming packet based on packet’s destination address Route lookup –Find the longest prefix in the table that matches the packet destination address INF5050, University Of Oslo41

42. IP Routing Packet with destination address 12.82.100.101 is sent to interface 2, as 12.82.100.xxx is the longest prefix matching packet ’ s destination address …… 312.82.xxx.xxx 1128.16.120.xxx 1 2 128.16.120.111 12.82.100.101 12.82.100.xxx 2 INF5050, University Of Oslo42

43. Patricia Tries P1111*H1 P210*H2 P31010*H3 P410101H4 P2 P3 P4 P1 A B C G D F H E 1 0 0 1 1 1 1 Lookup 10111 Add P5=1110* I 0 P5 next-hop-ptr (if prefix) left-ptr right-ptr Trie node INF5050, University Of Oslo43 Use binary tree paths to encode prefixes * The lookup operation determines if a string exists in a trie

44. Patricia Tries Properties Supported operations –Insertion, deletion, and searching Advantage –Simple to implement Disadvantage –One lookup may take O(m), where m is number of bits (32 in the case of IPv4) INF5050, University Of Oslo44

45. IP Route Lookup Schemes Cashing Address Trie-Based Schemes –Standard Trie Structure –Path-Compressed Trie –Multibit Trie –Level Compression Trie –Lulea Algorithm –Binary Search –Multiway Range Search –Two-Trie Structure Hardware-based schemes –DIR-24-8-BASIC Scheme –SRAM-Based Pipeline scheme –Ternary CAM (TCAM) INF5050, University Of Oslo45

46. Complexity Comparison INF5050, University Of Oslo46  N= Number of entries in the forwarding table  W= Length of the prefix  K = Number of bits * Remember that worst-case lookup time is proportional to W

47. Binary Trie Data structure of a 1-bit binary trie INF5050, University Of Oslo47

48. Forwarding Decisions ATM and MPLS switches –Direct Lookup Bridges and Ethernet switches –Associative Lookup –Hashing –Trees and tries IP Router Lookup and Lookup Schemes –Caching –Patricia trees/tries –Other methods Packet Classification INF5050, University Of Oslo48

49. Packet Classification INF5050, University Of Oslo Action ---- PredicateAction Classifier (Policy Database) Packet Classification Forwarding Engine Incoming Packet HEADERHEADER 49

50. Packet Classification Packet classification –Map each packet to a predefined flow/connection (for datagram forwarding) –Used to implement more sophisticated services (e.g., QoS) Flow –A subset of packets between any two endpoints in the network 1 2 Scheduler flow 1 flow 2 flow n Classifier Buffer management * Flows are specified by rules * Classifier - Set of rules INF5050, University Of Oslo50

51. Continued …. Packet Classification Classify an IP packet based on a number of fields in the packet header, e.g., –Source/destination IP address (32 bits) –Source/destination port number (16 bits) –Type of service (TOS) byte (8 bits) –Type of protocol (8 bits) 1 2 Scheduler flow 1 flow 2 flow n Classifier Buffer management INF5050, University Of Oslo51 * In general fields are specified by range

52. Output Functions Buffer management –Decides when and which packet to drop Scheduler –Decides when and which packet to transmit –One FIFO queue per flow 1 2 Scheduler Buffer INF5050, University Of Oslo52 Goals of Scheduler Quality of Service (QoS) Protection (stop a flow from hogging the entire output link) Fast

53. Priority Scheduler Continued … Router Examples Priority scheduler –Packets in the highest priority queue are always served before the packets in lower priority queues Round robin –Packets are served in a round-robin fashion High priority Medium priority Low priority Priority Scheduler High priority Medium priority Low priority INF5050, University Of Oslo53

54. Packet classification Rules Examples of classification rules –Access-control in firewalls Deny all e-mail traffic from ISP-X to Y –Policy-based routing Route IP telephony traffic from X to Y via ATM –Differentiate quality of service Ensure that no more than 50 Mbps are injected from ISP-X INF5050, University Of Oslo54

55. Continued … Classifier Example Packet Filtering –Rule R1 blocks all telnet connections from outside into Net A Policy Routing –Rule R2 enables the router to forward all real-time traffic using real-time transport protocol (rtp) in the application layer from Net B to Net D through the ATM network Traffic Policing –Rule R3 limits the total transmission control protocol (tcp) traffic rate from Net C to Net B to 10 Mb/s INF5050, University Of Oslo55

56. Metrics for packet classification Performance metrics for packet classification –Search speed –Storage requirement –Scalability in classifier size –Scalability in the number of header fields –Update time –Flexibility in specification INF5050, University Of Oslo56

57. Packet Classification Schemes Schemes for packet Classification –Basic Data Structure –Geometric Algorithms –Heuristics –Hardware-Based Algorithms INF5050, University Of Oslo57 Details of each scheme are available on the paper

58. Agenda Background: Introduction Packet Lookup and Classification Switching Fabrics INF5050, University Of Oslo58

59. Switching Fabrics Input and Output Queueing Input Queueing –Scheduling algorithms –Combining input and output queues –Multicast traffic –Other non-blocking fabrics Switching –Transfer of packets from input ports to output ports –Solution of the contentions for output ports Queueing – Where to store Scheduling –What to transfer INF5050, University Of Oslo59

60. Interconnects Two basic techniques INF5050, University Of Oslo Input Queueing Output Queueing Usually a non-blocking switch fabric (e.g. crossbar) Usually a fast bus 60

61. Input Queueing Practical Algorithms Maximal Size Algorithms –iSLIP (iterative round robin matching with slip) –Wave Front Arbiter (WFA) –Parallel Iterative Matching (PIM) Maximal Weight Algorithms –Fair Access Round Robin (FARR) –Longest Port First (LPF) INF5050, University Of Oslo61

62. iSLIP Scheduling Algorithm iSLIP? –An algorithm that determines which ingress ports will send to which egress ports across the crossbar fabric in a input-queued switch –Three key-steps (next slide for explanation) Request Grant Accept INF5050, University Of Oslo62

63. iSLIP Scheme steps INF5050, University Of Oslo63 Key steps

64. Continued….iSLIP INF5050, University Of Oslo 1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4 Requests 1 2 3 4 1 2 3 4 Grant 1 2 3 4 1 2 3 4 Accept/Match 1 2 3 4 1 2 3 4 #1 #2 Round-Robin Selection 1 2 3 4 1 2 3 4 64

65. iSLIP Properties Achieves up to100% throughput for uniform traffic Characteristics –Simple –Provides high throughput –Starvation-free –Operates at high speed Because of its simplicity INF5050, University Of Oslo65

66. iSLIP Operations Grant phase –Each output selects the requesting input at the pointer, or the next input in round-robin order. It only updates its pointer if the grant is accepted Accept phase –Each input selects the granting output at the pointer, or the next output in round-robin order INF5050, University Of Oslo66

67. iSLIP Implementation INF5050, University Of Oslo67 More details on the paper Read the paper if you are interested

68. INF5050, University Of Oslo68