ITTC High-Performance Networking The University of Kansas EECS 881 Architecture and Topology

Transcription

1 High-Performance Networking The University of Kansas EECS 881 Architecture and Topology James P.G. Sterbenz Department of Electrical Engineering & Computer Science Information Technology & Telecommunications Research Center The University of Kansas rev James P.G. Sterbenz

2 Network Architecture and Topology Outline AT.1 Topology and geography AT.2 Scale AT.3 Resource tradeoffs HSN-AT-2

3 Network Architecture and Topology Outline application application session session transport transport network network network network link link link link end system node network node end system link node network AT.1. Topology and geography AT.2. Scale AT.3. Resource Tradeoffs HSN-AT-3

4 Network Architecture and Topology Topology and Geography AT.1 Topology and geography AT.1.1 Scalability AT.1.2 Latency AT.1.3 Bandwidth AT.1.4 Virtual overlays and lightpaths AT.1.5 Practical constraints AT.2 Scale AT.3 Resource tradeoffs HSN-AT-4

5 infinite bandwidth CPU Ideal Network Model Network Path Principle R = network CPU M app end system M app end system D = 0 zero latency Network Path Principle The network must provide high-bandwidth low-latency paths between end systems. N-II HSN-AT-5

6 Link Topologies Shared Medium Shared medium (e.g. bus topology) problems? HSN-AT-6

7 Link Topologies Shared Medium Shared medium (e.g. bus topology) topology constrained by medium longest path limited to LAN lengths due to delay HSN-AT-7

8 Link Topologies Shared Medium Shared medium (e.g. bus topology) topology constrained by medium longest path limited to LAN lengths due to delay Network scalability? HSN-AT-8

9 Link Topologies Shared Medium Shared medium (e.g. bus topology) topology constrained by medium longest path limited to LAN lengths due to delay Limited scalability all hosts share capacity of bus HSN-AT-9

10 Link Topologies Shared Medium Shared medium (e.g. bus topology) topology constrained by medium longest path limited to LAN lengths due to delay Limited scalability all hosts share capacity of bus contention for shared medium HSN-AT-10

11 Link Topologies Shared Medium Shared medium (e.g. bus topology) topology constrained by medium longest path limited to LAN lengths due to delay Limited scalability all hosts share capacity of bus contention for shared medium collisions in shared medium need MAC (medium access control) HSN-AT-11

12 Link Topologies Shared Medium Shared medium (e.g. bus topology) topology constrained by medium longest path limited to LAN lengths due to delay Limited scalability all hosts share capacity of bus contention for shared medium Example collisions in shared medium need MAC (medium access control) original DIX Ethernet HSN-AT-12

13 Link Topologies Point-to-Point Mesh Point-to-point links between switches space division mesh HSN-AT-13

14 Link Topologies Point-to-Point Mesh Point-to-point links between switches space division mesh Network scalability? HSN-AT-14

15 Link Topologies Point-to-Point Mesh Point-to-point links between switches space division mesh Scalable: add links and expand switches HSN-AT-15

16 Link Topologies Point-to-Point Mesh Point-to-point links between switches space division mesh Scalable: add links and expand switches add switches HSN-AT-16

17 Scalable Topologies Mesh vs. Shared Medium Scalability of Mesh Topologies Mesh network technologies scale better then shared medium. Use power control and directional antennæ to increase spatial reuse in shared medium wireless networks. N-II.4 HSN-AT-17

18 Latency Network Path d i short path long path D = d i Network Latency Principle N-1Al The latency along a path is the sum of all its components. The benefit of optimising an individual link is directly proportional to its relative contribution to the end-to-end latency. HSN-AT-18

19 Latency Topology and Geography Constituents of network latency D = (h 1)d s + h d i Geography: speed of light delay d i SAN LAN MAN WAN GEO Mars Dia 100 m 1 km 100 km km km km RTT 1 μs 10 μs 1 ms 200 ms 480 ms 6 45 min Topology: number of hops h; switching delay d s Network Diameter Principle N-1Ah The number of per hop latencies along a path is bounded by the diameter of the network. The network topology should keep the diameter low. HSN-AT-19

20 Bandwidth Network Path bottleneck r i long path R = min(r i ) Network Bandwidth Principle The maximum bandwidth along a path is limited by the minimum bandwidth link or node, which is the bottleneck. There is no point in optimising a link that is not a bottleneck. N-1Ab HSN-AT-20

21 Overlay Networks Abstract and Hide Underlying Network Overlay networks hide underlying topology VPNs (virtual private networks) secure overlay sessions datagram overlay meshes lightpaths Network Overlay Principle N-IIo Overlay networks must provides the same high-performance paths as he physical networks. The number of overlay layers should be kept as small as possible, and overlays must be adaptable based on end-to-end path requirements and topology information form the lower layers. HSN-AT-21

22 Overlay Networks Lightpath Assignment Lightpath assignment problem preserve high-performance along overlay such that no link carries more than one light path of given wavelength HSN-AT-22

23 Network Architecture and Topology Network Scale AT.1 Topology and geography AT.2 Scale AT.2.1 Network engineering AT.2.2 Hierarchy AT.2.3 Bandwidth aggregation and isolation AT.2.4 Latency optimisation AT.2.5 Wireless network density AT.2.5 Practical constraints AT.3 Resource tradeoffs HSN-AT-23

24 Network Scale Network Engineering Parameters sparse dense topology degree # nodes diameter aggregation sparsely connected low high high low densely connected high low low high Network Scaling Principle N-5Ct Networks should be able to scale in size while balancing switch cost against hop count. HSN-AT-24

25 Network Scale Hierarchy Hierarchy important technique to control latency and aggregation A 1 B C bound state maintained divide network into clusters 4 4 state aggregated per cluster examples: Nimrod, P-NNI Network Hierarchy Principle Use hierarchy and clustering to manage network scale and complexity, and reduce the overhead of routing algorithms. N-5C HSN-AT-25

26 Network Scale Aggregation, Isolation, Latency Backbone WAN regional net MAN LAN neighborhood personal net home DAN Physical hierarchy to limit number of hops and control aggregation example: Internet (NSFNET) prior to 1994 HSN-AT-26

27 Network Scale Aggregation, Isolation, Latency Hierarchy: bandwidth isolation isolate bandwidth in different subnet layers exploit locality at network edge e.g. within campus or enterprise network aggregate bandwidth in network core Hierarchy to Aggregate and Isolate Bandwidth Use hierarchy to manage bandwidth aggregation in the core of the network, and to isolate clusters of traffic locality from one another. N-5B HSN-AT-27

28 Network Scale Aggregation, Isolation, Latency Hierarchy: latency engineering control network diameter and latency engineer low diameter at each level shallow hierarchy 3 5 levels home network or LAN access or campus network MAN WAN backbone Hierarchy to Minimise Latency Use hierarchy and cluster size to minimise network diameter and resultant latency. N-5Cl HSN-AT-28

29 Network Scale Wireless Network Density no control power control overlay Transmission power gives tradeoff between transmission range degree of connectivity Wireless Density Control to Optimise Degree and Diameter N-5Cw Use density control and long link overlays to optimise the tradeoff between dense low-diameter and sparse high-diameter wireless networks. HSN-AT-29

30 Policy-based routing Network Scale Practical Constraints Network provider deployment complex topologies peering not high-performance many hops/pop Administrative Constraints Increase the Importance of Good Design Backbone A Backbone W Backbone S MAN B MAN V 2 1 N-III.2 Policy and administrative constraints distort the criteria that govern the application of many high-performance network design principles. The importance of good (principled) design is elevated when these constraints are present. HSN-AT-30

31 Network Architecture and Topology Resource Tradeoffs AT.1 Topology and geography AT.2 Scale AT.3 Resource tradeoffs AT.3.1 Bandwidth, processing, and memory AT.3.2 Latency as a constraint AT.3.3 Relative scaling with high speed AT.2.4 Active networking HSN-AT-31

32 Resource Tradeoffs Resources and Constraints Network is a collection of resources and constraints: P processing M memory B bandwidth E energy or power L latency $ monetary cost Relative composition must be balanced to optimise cost and performance determine topology, engineering, and functional placement Network Resource Tradeoff & Engineering Principle N-2 Networks are collections of resources. The relative composition of these resources must be balanced to optimise cost and performance, and to determine network topology, engineering, and functional placement. HSN-AT-32

33 Resource Tradeoffs Resources and Constraints Network is a collection of resources and constraints: P processing M memory B bandwidth E energy or power L latency $ monetary cost Relative composition must be balanced to optimise cost and performance determine topology, engineering, and functional placement Objective function to determine optimal mix ( (P), (B), (M), (E), (L)) HSN-AT-33

34 Example 3.5 Example: content cache location Resource Tradeoffs $(B) =, $(M) = 0 where goes the content? Content Cache Location HSN-AT-34

35 Example 3.5 Resource Tradeoffs Content Cache Location Example: content cache location $(B) =, $(M) = 0 everyone has local copy of everything $(B) = 0, $(M) = where goes the content? HSN-AT-35

36 Example 3.5 Resource Tradeoffs Content Cache Location Example: content cache location $(B) =, $(M) = 0 everyone has local copy of everything $(B) = 0, $(M) = single copy on server: stream every time demanded HSN-AT-36

37 Example 3.5 Resource Tradeoffs Content Cache Location Example: content cache location $(B) =, $(M) = 0 everyone has local copy of everything $(B) = 0, $(M) = single copy on server: stream every time demanded is this realistic? HSN-AT-37

38 Example 3.5 Resource Tradeoffs Content Cache Location Example: content cache location $(B) =, $(M) = 0 everyone has local copy of everything $(B) = 0, $(M) = single copy on server: stream every time demanded M and B both have finite (and bounded cost): solution? HSN-AT-38

39 Example 3.5 Resource Tradeoffs Content Cache Location Example: content cache location plot cost of bandwidth vs. depth in network distribution tree cost bandwidth root normalized depth leaves HSN-AT-39

40 Example 3.5 Resource Tradeoffs Content Cache Location Example: content cache location plot cost of bandwidth vs. depth in network distribution tree plot cost of memory cost memory bandwidth root normalized depth leaves HSN-AT-40

41 Example 3.5 Resource Tradeoffs Content Cache Location Example: content cache location plot cost of bandwidth plot cost of memory what next? cost memory bandwidth root normalized depth leaves HSN-AT-41

42 Example 3.5 Resource Tradeoffs Content Cache Location Example: content cache location plot cost of bandwidth plot cost of memory l opt = min[ (B) + (M)] optimal level l cost memory bandwidth root normalized depth l opt leaves HSN-AT-42

43 Example 3.5 Resource Tradeoffs Content Cache Location Example: content cache location plot cost of bandwidth plot cost of memory l opt = min[ (B) + (M)] optimal level l latency constrains level maximum distance from client cost memory l opt = max(l Lmax, l opt ) bandwidth root normalized depth l opt l lat leaves HSN-AT-43

44 Resource Tradeoffs Relative Scaling Relative scaling of resources is important if all resources/constraints scale uniformly, no change speed of light remains constant latency becomes relatively more important bandwidth- -delay product requires increasing M technology scales non-uniformly example: Moore s law increase in P enables IP lookup/classification in hardware, active networking Resource Tradeoffs Change and Enable New Paradigms N-2A The relative cost of resources and constraints changes over time due to non-uniform advances in different aspects of technology. This should motivate constant rethinking about the way in which networks are structured and used. HSN-AT-44

45 Resource Tradeoffs Active Networking Decreasing cost of processing and memory enables more computation in the network nodes Active networking strong AN users inject capsules of code executed on nodes moderate AN network service providers provisions protocols and services may be dynamically provisioned with active packets HSN-AT-45

46 High-Performance Networking Acknowledgements The material in these foils comes from the textbook supplementary materials: Sterbenz & Touch, High-Speed Networking: A Systematic Approach to High-Bandwidth Low-Latency Communication HSN-AT-46