Overlay Networks EECS 122 Lecture 18 Department of

Overlay Networks EECS 122: Lecture 18 Department of Electrical Engineering and Computer Sciences University of California Berkeley April 3, 2003 A. Parekh, EE 122 S 2003. Revised and enhanced F'02 Lectures

What is an overlay network? n n n A network defined over another set of networks The overlay addresses its own nodes Links on one layer are network segments of lower layers q n A A’ A Requires lower layer routing to be utilized Overlaying mechanism is called tunneling April 3, 2003 A. Parekh, EE 122 S 2003. Revised and enhanced F'02 Lectures 2

Overlay Concept: Going Up C 5 4 7 8 6 11 2 A 1 10 3 13 12 B Overlay Network Nodes April 3, 2003 A. Parekh, EE 122 S 2003. Revised and enhanced F'02 Lectures 3

Overlay Concept: Going Up C 5 4 7 8 6 11 2 A 1 n n n 10 3 13 12 B Overlay Networks are extremely popular MBONE, Akamai, Virtual Private Networks, Napster, Gnutella Overlay Networks may even peer! April 3, 2003 A. Parekh, EE 122 S 2003. Revised and enhanced F'02 Lectures 4

Overlay Concept: Going Down 5 4 7 8 6 11 2 1 April 3, 2003 10 3 13 A. Parekh, EE 122 S 2003. Revised and enhanced F'02 Lectures 12 5

IP Network is the Overlay… 5 4 7 8 6 11 2 10 c 3 a 1 13 d 12 b ATM links can be the “physical layer” for IP April 3, 2003 A. Parekh, EE 122 S 2003. Revised and enhanced F'02 Lectures 6

IP Network is the overlay 5 4 7 8 6 11 2 10 c 1 3 a 13 d 12 b Virtual Circuit under Datagram! April 3, 2003 A. Parekh, EE 122 S 2003. Revised and enhanced F'02 Lectures 7

Example: Napster m 5 E m 6 F E? E E? m 5 m 1 m 2 m 3 m 4 m 5 m 6 April 3, 2003 m 4 C A m 1 D A B C D E F B m 3 m 2 A. Parekh, EE 122 S 2003. Revised and enhanced F'02 Lectures 8

Routing On the overlay Underlying Network April 3, 2003 A. Parekh, EE 122 S 2003. Revised and enhanced F'02 Lectures 9

Routing on the Overlay n The underlying network induces a complete graph of connectivity q No routing required! Underlying Network April 3, 2003 A. Parekh, EE 122 S 2003. Revised and enhanced F'02 Lectures 10

Routing on the Overlay n q 10 n 100 The underlying network induces a complete graph of connectivity 200 But q 90 90 10 q 100 20 No routing required! 100 q One virtual hop may be many underlying hops away. Latency and cost vary significantly over the virtual links State information may grow with E (n^2) 10 April 3, 2003 A. Parekh, EE 122 S 2003. Revised and enhanced F'02 Lectures 11

Routing Issues n The underlying network induces a complete graph of connectivity q n But q Underlying Network April 3, 2003 No routing required! q q One virtual hop may be many underlying hops away. Latency and cost vary significantly over the virtual links State information may grow with E (n^2) A. Parekh, EE 122 S 2003. Revised and enhanced F'02 Lectures 12

Routing Issues n 1 2 The underlying network induces a complete graph of connectivity q n But q 3 q q 4 April 3, 2003 5 No routing required! One virtual hop may be many underlying hops away. Latency and cost vary significantly over the virtual links State information may grow with E (n^2) A. Parekh, EE 122 S 2003. Revised and enhanced F'02 Lectures 13

Relating the virtual topology to the underlying network 1 2 4 5 3 Message from 4 1 April 3, 2003 A. Parekh, EE 122 S 2003. Revised and enhanced F'02 Lectures 14

Relating the virtual topology to the underlying network 1 2 3 1 2 4 5 3 4 5 Message from 4 1 4 3 2 1 April 3, 2003 4 1 3 1 5 2 5 1 Extreme Inefficiencies Possible A. Parekh, EE 122 S 2003. Revised and enhanced F'02 Lectures 15

Routing Issues n 1 2 The underlying network induces a complete graph of connectivity q n But q q 3 q 4 5 n One virtual hop may be many underlying hops away. Latency and cost vary significantly over the virtual links State information may grow with E (n^2) At any given time, the overlay network picks a connected subgraph based on nearest neighbors q q April 3, 2003 No routing required! How often can vary Also, structured (Chord) v/s unstructured (Gnutella) A. Parekh, EE 122 S 2003. Revised and enhanced F'02 Lectures 16

Kinds of Overlay Networks Three kinds of Overlays n 1. Only Hosts: Peer to Peer Networks (P 2 P) q 2. Example: Gnutella, Napster Only Gateway nodes: Infrastructure Overlays q Content Distribution Networks (CDNs) q 3. Example: Akamai Host and Gateway Nodes: q Virtual Private Networks Overlay node structure n q q Regular: Chord, Pastry Adhoc: Gnutella Functions n q q Route Enhancement: Better Qo. S, Application Level Multicast Resource Discovery: P 2 P April 3, 2003 A. Parekh, EE 122 S 2003. Revised and enhanced F'02 Lectures 17

Outline n Infrastructure Overlays q q n P 2 P Overlays q n Application Level Multicast Example of a P 2 P Overlay q n Resource Discovery in Gnutella Example of an Infrastructure Overlay q n Adding performance and route functionality Resource Discovery Content Addressable Networks Conclusions April 3, 2003 A. Parekh, EE 122 S 2003. Revised and enhanced F'02 Lectures 18

Infrastructure Overlays n Overlay network users are not directly connected to the overlay nodes n 1 2 4 5 E. g. Akamai 3 April 3, 2003 A. Parekh, EE 122 S 2003. Revised and enhanced F'02 Lectures 19

Overlay Routing: Edge Mapping n 1 2 Overlay network users are not directly connected to the overlay nodes n n n IP(5) 3 User must be redirected to a “close by” overlay node Edge-Mapping, or redirection function is hard since q ? 4 5 q n # potential users enormous User clients not under direct control When overlay clients are directly connected the edge mapping function is obviated q April 3, 2003 E. g. Akamai A. Parekh, EE 122 S 2003. Revised and enhanced F'02 Lectures E. g. P 2 P: users/nodes colocated 20

Overlay Routing: Edge Mapping n 1 n 2 Overlay nodes interconnect clients Enhance nature of connection q q q n IP(5) 3 Multicast Secure Low Loss Much easier to add functionality than to integrate into a router ? 4 April 3, 2003 5 A. Parekh, EE 122 S 2003. Revised and enhanced F'02 Lectures 21

Overlay Routing: Adding Function to the route n 1 2 n Overlay nodes interconnect clients Enhance nature of connection q q q n 3 n 4 April 3, 2003 Multicast Secure Low Loss Much easier to add functionality than to integrate into a router Overlay nodes can become bottlenecks 5 A. Parekh, EE 122 S 2003. Revised and enhanced F'02 Lectures 22

Overlay Routing: Resource Location n Overlay network may contain resources. Eg. q 1 2 q n B D E B F n 3 B? 4 April 3, 2003 A B C Servers Files Client makes request for resource Overlay must “search” for “closest” node that has the resource q 5 E. g. find the least loaded server that has a piece of content and that is has low network latency to client A D A. Parekh, EE 122 S 2003. Revised and enhanced F'02 Lectures 23

Overlay Routing: Resource Location n D B 1 q 2 q B B F n D E F C n 3 A C D E 4 A B C n n April 3, 2003 Servers Files Client makes request for resource Overlay must “search” for “closest” node that has the resource q B? 5 A D Overlay network may contain resources. Eg. E. g. find the least loaded server that has a piece of content and that is has low network latency to client A single “index” is not scalable Overlay launches a query to locate resource A. Parekh, EE 122 S 2003. Revised and enhanced F'02 Lectures 24

Overlay Routing: Resource Location n D B 1 q 2 q B B F n D E F C n 3 A C D E 4 A B C n n n April 3, 2003 Servers Files Client makes request for resource Overlay must “search” for “closest” node that has the resource q B? 5 A D Overlay network may contain resources. Eg. E. g. find the least loaded server that has a piece of content and that is has low network latency to client A single “index” is not scalable Overlay launches a query to locate resource Query is “Routed” through the overlay until object is located A. Parekh, EE 122 S 2003. Revised and enhanced F'02 Lectures 25

Overlay Routing: Resource Location n D B 1 q 2 q B B F n D E 4 F C 3 n 4 A C D E 4 A B C 5 A D q 4 n n Servers Files Client makes request for resource Overlay must “search” for “closest” node that has the resource B? n April 3, 2003 Overlay network may contain resources. Eg. E. g. find the least loaded server that has a piece of content and that is has low network latency to client A single “index” is not scalable Overlay launches a query to locate resource Query is “Routed” through the overlay until object is located A. Parekh, EE 122 S 2003. Revised and enhanced F'02 Lectures 26

Overlay Routing: Resource Location n D B 1 q 2 q B B F n D E 4 F C 3 n 4 A C D E 4 A B C 5 A D q 4 n n Servers Files Client makes request for resource Overlay must “search” for “closest” node that has the resource B? n April 3, 2003 Overlay network may contain resources. Eg. E. g. find the least loaded server that has a piece of content and that is has low network latency to client A single “index” is not scalable Overlay launches a query to locate resource Query is “Routed” through the overlay until object is located A. Parekh, EE 122 S 2003. Revised and enhanced F'02 Lectures 27

P 2 P Overlays n Overlay network users are not directly connected to the overlay nodes n E. g. Napster, Gnutella No edge mapping problem n No gateways to maintain But n Nodes have limited resources n q q q April 3, 2003 A. Parekh, EE 122 S 2003. Revised and enhanced F'02 Lectures storage, connectivity computational power 28

Gnutella n n n Distribute file location Idea: multicast the request Hot to find a file: q Send request to all neighbors q Neighbors recursively multicast the request q Eventually a machine that has the file receives the request, and it sends back the answer Advantages: q Totally decentralized, highly robust Disadvantages: q Not scalable; the entire network can be swamped with request (to alleviate this problem, each request has a TTL) April 3, 2003 A. Parekh, EE 122 S 2003. Revised and enhanced F'02 Lectures 29

Gnutella: Example n Assume: m 1’s neighbors are m 2 and m 3; m 3’s neighbors are m 4 and m 5; … m 5 E m 6 F E E? D E? m 4 E? E? C A m 1 April 3, 2003 B m 3 m 2 A. Parekh, EE 122 S 2003. Revised and enhanced F'02 Lectures 30

Summary n Two kinds of overlays functions q q n Two kinds of virtual topologies q q n Structured: mesh, ring etc. Unstructured Two kinds of client connectivty q q n Overlay provides access to distributed resources Overlay facilitates communication among other client applications Direct: P 2 P Not direct: Akamai Overlay Network Functions q q Select Virtual Edges (fast or slow timescales) Overlay Routing Protocol Edge Mapping Resource Location April 3, 2003 A. Parekh, EE 122 S 2003. Revised and enhanced F'02 Lectures 31

Example: Application Level Multicast Content Producer Media Distribution Network Media Clients April 3, 2003 A. Parekh, EE 122 S 2003. Revised and enhanced F'02 Lectures Content Producer 32

The Broadcast Internet Content Producer April 3, 2003 A. Parekh, EE 122 S 2003. Revised and enhanced F'02 Lectures 33

Broadcast Overlay Architecture Management Platform content management Media Delivery System Redirection April 3, 2003 A. Parekh, EE 122 S 2003. Revised and enhanced F'02 Lectures injection & real-time control network management monitoring & provisioning server management redirection management load balancing system availability viewer management subscriptions, PPV, monitoring, Neilson ratings, targeted advertising 34

Broadcast Management n n n April 3, 2003 A. Parekh, EE 122 S 2003. Revised and enhanced F'02 Lectures Application-level information for management and tracking Works across multiple networks Content Producer event programming with ad-hoc query audience statistics 35

Broadcast Manager Node Information Stream Switchover April 3, 2003 A. Parekh, EE 122 S 2003. Revised and enhanced F'02 Lectures 36

Policy Management April 3, 2003 A. Parekh, EE 122 S 2003. Revised and enhanced F'02 Lectures 37

Example: Content Addressable P 2 P Networks (CAN) n CAN is one of several recent P 2 P architectures that q q imposes a structure on the virtual topology uses a distributed hash-table data structure abstraction n q q q n Note: item can be anything: a data object, document, file, pointer to a file… routes queries through the structured overlay attempts to distribute (object, location) pairs uniformly throughout the network supports object lookup, insertion and deletion of objects efficiently. Others: Chord, Pastry, Tapestry April 3, 2003 A. Parekh, EE 122 S 2003. Revised and enhanced F'02 Lectures 38

Content Addressable Network (CAN) Associate to each node and item a unique id in an d-dimensional space n Properties n Routing table size O(d) q Guarantee that a file is found in at most d*n 1/d steps, where n is the total number of nodes q April 3, 2003 A. Parekh, EE 122 S 2003. Revised and enhanced F'02 Lectures 39

CAN Example: Two Dimensional Space n n Space divided between nodes All nodes cover the entire space Each node covers either a square or a rectangular area of ratios 1: 2 or 2: 1 Example: Assume space size (8 x 8) q Node n 1: (1, 2) first node that joins cover the entire space q 7 6 5 4 3 n 1 2 1 0 0 April 3, 2003 A. Parekh, EE 122 S 2003. Revised and enhanced F'02 Lectures 1 2 3 4 5 6 7 40

CAN Example: Two Dimensional Space n Node n 2: (4, 2) joins space is divided between n 1 and n 2 7 6 5 4 3 n 2 n 1 2 1 0 0 April 3, 2003 A. Parekh, EE 122 S 2003. Revised and enhanced F'02 Lectures 1 2 3 4 5 6 7 41

CAN Example: Two Dimensional Space n Node n 2: (4, 2) joins space is divided between n 1 and n 2 7 6 n 3 5 4 3 n 2 n 1 2 1 0 0 April 3, 2003 A. Parekh, EE 122 S 2003. Revised and enhanced F'02 Lectures 1 2 3 4 5 6 7 42

CAN Example: Two Dimensional Space n Nodes n 4: (5, 5) and n 5: (6, 6) join 7 6 n 5 n 4 n 3 5 4 3 n 2 n 1 2 1 0 0 April 3, 2003 A. Parekh, EE 122 S 2003. Revised and enhanced F'02 Lectures 1 2 3 4 5 6 7 43

CAN Example: Two Dimensional Space n n Nodes: n 1: (1, 2); n 2: (4, 2); n 3: (3, 5); n 4: (5, 5); n 5: (6, 6) Items: f 1: (2, 3); f 2: (5, 1); f 3: (2, 1); f 4: (7, 5); 7 6 n 5 n 4 n 3 5 f 4 4 f 1 3 n 2 n 1 2 f 3 1 f 2 0 0 April 3, 2003 A. Parekh, EE 122 S 2003. Revised and enhanced F'02 Lectures 1 2 3 4 5 6 7 44

CAN Example: Two Dimensional Space n Each item is stored by the node who owns its mapping in the space 7 6 n 5 n 4 n 3 5 f 4 4 f 1 3 n 2 n 1 2 f 3 1 f 2 0 0 April 3, 2003 A. Parekh, EE 122 S 2003. Revised and enhanced F'02 Lectures 1 2 3 4 5 6 7 45

CAN: Query Example n n n Each node knows its neighbors in the d-space 7 Forward query to the neighbor that is closest to the 6 query id 5 Example: assume n 1 queries 4 f 4 n 5 n 4 n 3 f 4 f 1 3 n 2 n 1 2 f 3 1 f 2 0 0 April 3, 2003 A. Parekh, EE 122 S 2003. Revised and enhanced F'02 Lectures 1 2 3 4 5 6 7 46

Adding/Deleting nodes n n n April 3, 2003 New node picks a point P at random Assuming it can find any overlay node, it sends a join message to the node which owns that point When the message has reached P, the node divides itself in half along one of the dimensions (first x then y etc) Pairs are transferred and neighbor sets updated Similar reasoning handles departures and failures A. Parekh, EE 122 S 2003. Revised and enhanced F'02 Lectures 47

Relating Virtual Topology to the Underlying Network n n n Example: Three landmarks q q q n n n 0 -30 ms: level 0 31 -100 ms: level 1 101 -300 ms: level 2 Node j measures latencies of 10 ms, 110 ms, 40 ms to the three landmarks. The bin of node j is q n Neighbors should be close to each other in terms of latency on the underlying network Pick a set of well known landmark hosts Each node distributively computes its “bin” q q (l 1, l 3, l 2 : 021) q April 3, 2003 Nodes in the same bin are “close” to each other Orders the landmark set in increasing order of RTT from it. Latency is partitioned into levels Thus, associated with each landmark, at each node is a rank and a level. These values identify the bin A. Parekh, EE 122 S 2003. Revised and enhanced F'02 Lectures 48

Your standard Networking Functions… n Addressing: Uniquely identify the nodes q q n Topology Update: Characterize and maintain connectivity q q q n n Discover topology Measure “distance” metric(s) Dynamically provision (on slower timescale) Destination Discovery: Find node identifiers of the destination set Route Computation: Pick the tree (path) q q n host IP address, group address, attributes set is dynamic! Kind of path: Multicast, Unicast Global or Distributed Algorithm Policy Hierarchy Switching: Forward the packets at each node April 3, 2003 A. Parekh, EE 122 S 2003. Revised and enhanced F'02 Lectures 49

And Their Overlay Analogs n Addressing: Uniquely identify the nodes q q n Topology Update: Characterize and maintain connectivity q q q n n Discover topology Measure “distance” metric(s) Add/Insert Nodes, Binning Dynamically provision (on slower timescale) Destination Discovery: Find node identifiers destination Resource Location of the Edge Mapping set Route Computation: Pick the tree (path) q q n host IP address, group address, attributes set. Structured is dynamic!Topology Kind of path: Multicast, Unicast Application Level Routing. E. . g streaming broadcast Global or Distributed Algorithm Structured Topology Policy Hierarchy Switching: Forward the packets at each node April 3, 2003 A. Parekh, EE 122 S 2003. Revised and enhanced F'02 Lectures 50

Conclusions n n Overlays are an irreversible trend in network Overlays add new functions to the network infrastructure much faster than q q n Disadvantages q q n by trying to integrate them in the router relying on a infrastructure service provider on deploy the function Overlay nodes can create performance bottlenecks New end-to-end protocols may not work since the overlay nodes don’t understand them Generally better to improve performance by building an “underlay” and add functionality by building an overlay April 3, 2003 A. Parekh, EE 122 S 2003. Revised and enhanced F'02 Lectures 51