L13 P2P Overlay Networks: Theory

Transcription

1 L13 P2P Overlay Networks: Theory by T.S.R.K. Prasad EA C451 Internetworking Technologies

2 References / Acknowledgements Ch 6: Peer-to-Peer Content Networks, [Hoffman] Sec 2.6: Peer-to-Peer Applications, [Kurose] [Theotokis] Stephanos Androutsellis-Theotokis, A Survey of Peer-to-Peer File Sharing Technologies References

3 References / Acknowledgements [Ross-P2P] Keith W. Ross, and Dan Rubenstein, P2P Systems, Infocomm Tutorial [Kurose-P2P] Ch2: Applications, [Kurose] [Zhang-P2P] Prof. Zhi-Li Zhang, P2P, CSci5221: Foundations of Advanced Networking, Spring ( 2011/csci5221/index.php) [Rexford-P2P] Jennifer Rexford, Peer-to-Peer File Sharing, COS 461: Computer Networks ( References

4 References / Acknowledgements [Stoica-P2P] Ion Stoica, P2P Networks, EE122: Computer Networking, Dept of EECS, UCB, Fall (www-inst.eecs.berkeley.edu/~ee122/fa02) References

5 Optional Readings [Andersen] David G. Andersen, Hari Balakrishnan, M. Frans Kaashoek, and Robert Morris, Resilient Overlay Networks, SOPS [Lua] Eng Keong Lua, Jon Crowcroft, Marcelo Pias, Ravi Sharma and Steven Lim, A Survey and Comparison of Peer-to-Peer Overlay Network Schemes, IEEE Communications Survey and Tutorial, March [Rodrigues] Rodrigo Rodrigues, and Peter Druschel, Peer-to- Peer Systems Optional Reading

6 Presentation Overview Structured P2P Networks Unstructured P2P Networks Overlay Networks File Distribution Example Introduction Lecture Outline

7 Presentation Overview Structured P2P Networks Unstructured P2P Networks Overlay Networks File Distribution Example Introduction Lecture Outline

8 Client-Server Limitations Scalability is hard to achieve Presents a single point of failure Requires administration Unused resources at the network edge P2P systems try to address these limitations Introduction C/S Limitations

9 Defintion of P2P 1) Significant autonomy from central servers 2) Exploits resources at the edges of the Internet storage and content CPU cycles human presence 3) Resources at edge have intermittent connectivity, being added & removed Introduction Definition of P2P

10 It s a broad definition: P2P file sharing Napster, Gnutella, KaZaA, etc P2P communication Instant messaging P2P computation seti@home DHTs & their apps Chord, CAN, Pastry, Tapestry P2P apps built over emerging overlays PlanetLab Introduction Definition of P2P

11 Pure P2P architecture no always-on server arbitrary end systems directly communicate peers are intermittently connected and change IP addresses Introduction Pure P2P Architecture

12 Characteristics of P2P Networks Clients are also servers and routers Nodes contribute content, storage, memory, CPU Nodes are autonomous (no administrative authority) Network is dynamic: nodes enter and leave the network frequently Nodes collaborate directly with each other (not through well-known servers) Nodes have widely varying capabilities Introduction Characteristics of P2P Networks

13 Benefits of P2P Networks Efficient use of resources Unused bandwidth, storage, processing power at the edge of the network Scalability Consumers of resources also donate resources Aggregate resources grow naturally with utilization Introduction Benefits of P2P Networks

14 Benefits of P2P Networks Reliability Replicas Geographic distribution No single point of failure Ease of administration Nodes self organize No need to deploy servers to satisfy demand (c.f. scalability) Built-in fault tolerance, replication, and load balancing Introduction Benefits of P2P Networks

15 Key Issues in P2P Networks Join/leave How do nodes join/leave? Who is allowed? Search and retrieval How to find content? How are metadata indexes built, stored, distributed? Content Distribution Where is content stored? How is it downloaded and retrieved? Introduction Key Issues P2P Networks

16 Four Key Primitives (APIs) Join How to enter/leave the P2P system? Publish How to advertise a file? Search how to find a file? Fetch how to download a file? Introduction Four Key Primitives

17 Presentation Overview Structured P2P Networks Unstructured P2P Networks Overlay Networks File Distribution Example Introduction Lecture Outline

18 File Distribution: Client-Server vs Peer-to-Peer Question: how much time to distribute file (size F) from one server to N peers? peer upload/download capacity is limited resource F bits Internet File Distribution Example The Problem

19 Server Distributing a Large File F bits d 4 upload rate u s Internet d 3 d 1 d 2 upload rates u i download rates d i File Distribution Example Server Distributing a Large File

20 Server Distributing a Large File Sending an F-bit file to N receivers Transmitting NF bits at rate u s takes at least NF/u s time Receiving the data at the slowest receiver Slowest receiver has download rate d min = min i {d i } takes at least F/d min time Download time: max{nf/u s, F/d min } File Distribution Example Server Distributing a Large File

21 Speeding Up the File Distribution Increase the server upload rate Higher link bandwidth at the server Multiple servers, each with their own link Alternative: have the receivers help Receivers get a copy of the data and redistribute to other receivers To reduce the burden on the server File Distribution Example Speeding up the File Distribution

22 Peers Help Distributing a Large File F bits d 4 upload rate u s Internet u 4 d 3 d 1 u 1 u 2 u 3 d 2 upload rates u i download 22 rates d i File Distribution Example Peer Helpout

23 Peers Help Distributing a Large File Components of distribution latency Server must send each bit: min time F/u s Slowest peer must receive each bit: min time F/d min Upload time using all upload resources Total number of bits: NF Total upload bandwidth u s + sum i (u i ) Total: max{f/u s, F/d min, NF/(u s +sum i (u i ))} 23 File Distribution Example Peer Helpout

24 Peer-to-Peer is Self-Scaling Download time grows slowly with N Client-server: max{nf/u s, F/d min } Peer-to-peer: max{f/u s, F/d min, NF/(u s +sum i (u i ))} But Peers may come and go Peers need to find each other Peers need to be willing to help each other 24 File Distribution Example P2P is Self-Scaling

25 Minimum Distribution Time Client-server vs. P2P: example client upload rate = u, F/u = 1 hour, u s = 10u, d min u s P2P Client-Server N File Distribution Example Comparison

26 P2P file distribution: BitTorrent file divided into 256Kb chunks peers in torrent send/receive file chunks tracker: tracks peers participating in torrent torrent: group of peers exchanging chunks of a file Alice arrives obtains list of peers from tracker and begins exchanging file chunks with peers in torrent File Distribution Example BitTorrent

27 Presentation Overview Structured P2P Networks Unstructured P2P Networks Overlay Networks File Distribution Example Introduction Lecture Outline

28 Overlay networks overlay edge Overlay Networks

29 Overlay graph Virtual edge TCP connection or simply a pointer to an IP address Overlay maintenance Periodically ping to make sure neighbor is still alive Or verify liveness while messaging If neighbor goes down, may want to establish new edge New node needs to bootstrap Overlay Networks Overlay Graph

30 Overlays: all in the application layer Tremendous design flexibility Topology, maintenance Message types Protocol Messaging over TCP or UDP Underlying physical net is transparent to developer But some overlays exploit proximity application transport network data link physical application transport network data link physical application transport network data link physical Overlay Networks Application Layer

31 Overlay Networks Normal View Overlay Networks Normal View

32 Overlay Networks Focus at the application level Overlay Networks Application Level View

33 Examples of overlays DNS BGP routers and their peering relationships Content distribution networks (CDNs) Application-level multicast economical way around barriers to IP multicast And P2P apps! Overlay Networks Examples of Overlays

34 More about overlays Unstructured overlays e.g., new node randomly chooses three existing nodes as neighbors Structured overlays e.g., edges arranged in restrictive structure Proximity Not necessarily taken into account Overlay Networks More about Overlays

35 Presentation Overview Structured P2P Networks Unstructured P2P Networks Overlay Networks File Distribution Example Introduction Lecture Outline

36 Unstructured P2P Networks Central Server / Directory (Napster) Flooding (Gnutella) Super Peers / Super Nodes (KaZaA) Swarming (BitTorrent) Unstructured P2P Networks Types

37 Central Server / Directory Join and Publish File list and IP address is uploaded A centralized directory 1 1 C 1 B Unstructured P2P Networks Central Server

38 Central Server / Directory Search (a) User requests file search at server (b) Server responds with list of hosts (IP addresses) containing the file. centralized directory 2b 2a Unstructured P2P Networks Central Server

39 Central Server / Directory Fetch (a) User pings the sources that have the data (Looks for host with best bandwidth) centralized directory Pings Pings Unstructured P2P Networks Central Server

40 Central Server / Directory Fetch (a) User pings the sources that have the data (Looks for host with best bandwidth) (b) Downloads the file centralized directory Downloads the file Unstructured P2P Networks Central Server

41 Pros and Cons of Central Server Pros: Simple Search scope is O(1) Controllable (pro or con?) Cons: Server maintains O(N) State Server does all processing Single point of failure Unstructured P2P Networks Central Server

42 Unstructured P2P Networks Central Server / Directory (Napster) Flooding (Gnutella) Super Peers / Super Nodes (KaZaA) Swarming (BitTorrent) Unstructured P2P Networks Types

43 Query Flooding Join contact a few nodes to become neighbors Publish no need! Search ask neighbors, who ask their neighbors Fetch get file directly from another node Unstructured P2P Networks Flooding

44 Search by Flooding xyz.mp3 search xyz.mp3? Flooding Unstructured P2P Networks Flooding

45 Search by Flooding xyz.mp3 xyz.mp3? search Flooding Unstructured P2P Networks Flooding

46 Search by Flooding transfer Unstructured P2P Networks Flooding

47 Flooding: Pros and Cons Advantages Fully decentralized Search cost distributed Processing per node permits powerful search semantics Disadvantages Search scope is O(N) Search time may be quite long High overhead, and nodes come and go often Unstructured P2P Networks Flooding

48 Unstructured P2P Networks Central Server / Directory (Napster) Flooding (Gnutella) Super Peers / Super Nodes (KaZaA) Swarming (BitTorrent) Unstructured P2P Networks Types

49 Super Node Hierarchy Join on start, the client contacts a super-node Publish client sends list of files to its super-node Search queries flooded among super-nodes Fetch get file directly from one or more peers Unstructured P2P Networks Super Nodes Hierarchy

50 Motivation for Super Nodes Query consolidation Many connected nodes may have only a few files Propagating query to a sub-node may take more time than for the super-node to answer itself Stability Super-node selection favors nodes with high uptime How long you ve been on is a good predictor of how long you ll be around in the future Unstructured P2P Networks Super Nodes Motivation

51 Super Node Architecture Each peer is either a super node or is assigned to a super node Each super node knows about many other super nodes (almost mesh overlay) supernodes Unstructured P2P Networks Super Nodes Architecture

52 Super Node Architecture: After Join and Publish supernodes H 3 xyz.mp3 H 1 H 3 xyz.mp3 abc.mp3 H 2 xyz.mp3 H 1 H 2 abc.mp3 xyz.mp3 Unstructured P2P Networks Super Nodes After Join and Publish

53 Super Node Architecture: Search Query Flooding xyz.mp3 H 3 xyz.mp3? supernodes H 1 H 3 xyz.mp3 abc.mp3 H 2 xyz.mp3 H 1 H 2 abc.mp3 xyz.mp3 Unstructured P2P Networks Super Nodes Query Flooding

54 Super Node Architecture: Search Query Responses xyz.mp3 H 3 xyz.mp3 available at H 1, H 2, H 3 H 1 supernodes H 3 xyz.mp3 abc.mp3 H 2 xyz.mp3 H 1 H 2 abc.mp3 xyz.mp3 Unstructured P2P Networks Super Nodes Query Response

55 Super Node Architecture: Search Fetch xyz.mp3 H 3 supernodes H 1 H 3 xyz.mp3 abc.mp3 H 2 xyz.mp3 H 1 H 2 abc.mp3 xyz.mp3 Parallel download of pieces of file possible Unstructured P2P Networks Super Nodes Fetch

56 Pros and Cons of Super Nodes Pros: Tries to take into account node heterogeneity: Bandwidth Host Computational Resources Host Availability (?) Can take into account network locality Cons: Mechanisms easy to circumvent Still no real guarantees on search scope or search time Unstructured P2P Networks Super Nodes

57 Unstructured P2P Networks Central Server / Directory (Napster) Flooding (Gnutella) Super Peers / Super Nodes (KaZaA) Swarming (BitTorrent) Unstructured P2P Networks Types

58 Swarming Swarming: Download from others who are downloading the same object at the same time Join contact centralized tracker server, get a list of peers. Publish Run a tracker server. Search Out-of-band. E.g., use Google to find a tracker for the file you want. Fetch Download chunks of the file from your peers. Upload chunks you have to them. Unstructured P2P Networks Swarming

59 BitTorrent: Publish/Join Tracker Unstructured P2P Networks Swarming Publish / Join

60 BitTorrent: Fetch Unstructured P2P Networks Swarming Fetch

61 Presentation Overview Structured P2P Networks Unstructured P2P Networks Overlay Networks File Distribution Example Introduction Lecture Outline

62 The Search (Routing) Problem N 1 N 2 N 3 Key= title Value=MP3 data Publisher Internet? Client Lookup( title ) N 4 N 5 N 6 Structured P2P Networks The Search Problem

63 Routed Queries in Structured P2P Networks Publisher Key= title Value=MP3 data N 4 N 1 N 2 N 3 Client Lookup( title ) N 6 N 9 N 7 N 8 In structured overlay networks, searching is equivalent to routing on the structure of the overlay. Structured P2P Networks Routing in Structured P2P Networks

64 Distributed Hash Tables Academic answer to p2p Goals Guatanteed lookup success Provable bounds on search time Provable scalability Makes some things harder Fuzzy queries / full-text search / etc. Read-write, not read-only Hot Topic in networking since introduction in ~2000/2001 Structured P2P Networks DHT

65 DHT: Overview Abstraction: a distributed hash-table (DHT) data structure: put(id, item); item = get(id); Implementation: nodes in system form a distributed data structure Can be Ring, Tree, Hypercube, Skip List, Butterfly Network,... Structured P2P Networks DHT Overview

66 DHT: Overview Structured Overlay Routing: Join: On startup, contact a bootstrap node and integrate yourself into the distributed data structure; get a node id Publish: Route publication for file id toward a close node id along the data structure Search: Route a query for file id toward a close node id. Data structure guarantees that query will meet the publication. Fetch: Two options: Publication contains actual file => fetch from where query stops Publication says I have file X => query tells you has X, use IP routing to get X from Structured P2P Networks DHT Overview

67 DHT as Library / Platform Distributed application put(key, data) Distributed hash table get (key) data node node. node DHT provides the information look up service for P2P applications. Nodes uniformly distributed across key space Nodes form an overlay network Nodes maintain list of neighbors in routing table Decoupled from physical network topology Structured P2P Networks DHT as Library

68 DHT Schemes for Structured P2P Networks Chord Pastry Tapestry Content Addressable Network (CAN) Structured P2P Networks DHT Schemes

69 Chord Overview Associate to each node and file a unique id in an unidimensional space (a Ring) E.g., pick from the range [0...2 m ] Usually the hash of the file or IP address Properties: Routing table size is O(log N), where N is the total number of nodes Guarantees that a file is found in O(log N) hops from MIT in 2001 Structured P2P Networks Chord Overview

70 Chord IDs Key identifier = SHA-1(key) Node identifier = SHA-1(IP address) Both are uniformly distributed Both exist in the same ID space How to map key IDs to node IDs? The heart of Chord protocol is consistent hashing Structured P2P Networks Chord Chord IDs

71 Chord API insert(key, value)à store key/value at r nodes lookup(key) update(key, newval) join(n) leave() Structured P2P Networks Chord Chord API

72 DHT: Consistent Hashing Node 105 Key 5 K5 N105 K20 Circular ID space N32 N90 K80 A key is stored at its successor: node with next higher ID Structured P2P Networks Chord Consistent Hashing

73 DHT: Chord Basic Lookup N105 N120 N10 Where is key 80? N90 has K80 N32 K80 N90 N60 Structured P2P Networks Chord Basic Lookup

74 Basic Chord Lookup (Routing) Algorithm Lookup(my-id, key-id) n = my successor if my-id < n < key-id Lookup(id) on node n /go to next hop/ else return my successor /found the correct node/ Correctness depends only on successors O(N) lookup time, but we can do better Structured P2P Networks Chord Basic Lookup Algorithm

75 DHT: Chord Finger Table 1/4 1/2 1/8 1/16 1/32 1/64 1/128 N80 Entry i in the finger table of node n is the first node that succeeds or equals n + 2 i In other words, the i th finger points 1/2 n-i way around the ring Structured P2P Networks Chord Finger Table

76 DHT: Chord Join Assume an identifier space [0..8] Node n1 joins Succ. Table i id+2 i succ Structured P2P Networks Chord Join

77 DHT: Chord Join Node n2 joins Succ. Table i id+2 i succ Succ. Table i id+2 i succ Structured P2P Networks Chord Join

78 DHT: Chord Join Nodes n0, n6 join Succ. Table i id+2 i succ Succ. Table Succ. Table i id+2 i succ i id+2 i succ Succ. Table i id+2 i succ Structured P2P Networks Chord Join

79 DHT: Chord Join Nodes: n1, n2, n0, n6 Succ. Table i id+2 i succ Items 7 Items: f7, f2 0 Succ. Table 7 1 i id+2 i succ Succ. Table 6 2 i id+2 i succ Succ. Table i id+2 i succ Items 1 Structured P2P Networks Chord Join

80 DHT: Chord Routing Upon receiving a query for item id, a node: Checks whether stores the item locally If not, forwards the query to the largest node in its successor table that does not exceed id Succ. Table i id+2 i succ Succ. Table 7 1 i id+2 i succ query(7) Succ. Table i id+2 i succ Items 7 Succ. Table i id+2 i succ Items 1 Structured P2P Networks Chord Routing

81 Chord Routing Algorithm Lookup(my-id, key-id) look in local finger table for highest node n such that my-id < n < key-id if n exists Lookup(id) on node n /go to next hop/ else return my successor /found the correct node/ Structured P2P Networks Chord Routing Algorithm

82 Node Joining in Chord Node n joins the system: n picks a random identifier, id n performs n = lookup(id) n->successor = n n ->predecessor = n Structured P2P Networks Chord Node Joining

83 State Maintenance: Stabilization Protocol Periodically node n Asks its successor, n, about its predecessor n If n is between n and n n->successor = n notify n that n is its predecessor When node n receives notification message from n If n is between n ->predecessor and n, then n ->predecessor = n Improve robustness Each node maintain a successor list(usually of size 2*log N) Structured P2P Networks Chord State Maintenance

84 DHT: Network Locality in Chord Nodes close on a ring can be far on the network Structured P2P Networks Chord Network Locality

85 DHT: Chord Summary Routing table size? Log N fingers Routing time? Each hop expects to 1/2 the distance to the desired id => expect O(log N) hops. Structured P2P Networks Chord Summary

86 DHT Schemes for Structured P2P Networks Chord Pastry Tapestry Content Addressable Network (CAN) Structured P2P Networks DHT Schemes

87 Pastry: Overview Similar interface to Chord Considers network locality to minimize hops messages travel New node needs to know a nearby node to achieve locality Each routing hop matches the destination identifier by one more digit Many choices in each hop (locality possible) Structured P2P Networks Pastry Overview

88 Object Distribution O objid Consistent hashing [Karger et al. 97] 128 bit circular id space nodeids (uniform random) nodeids objids (uniform random) Invariant: node with numerically closest nodeid maintains object Structured P2P Networks Pastry Object Distribution

89 Object Insertion/Lookup O X Msg with key X is routed to live node with nodeid closest to X Route(X) Problem: complete routing table not feasible Structured P2P Networks Pastry Object Insertion/Lookup

90 Routing (Lookup) in Pastry d46a1c d471f1 d467c4 d462ba d4213f 65a1fc Route(d46a1c) d13da3 Properties log 16 N steps O(log N) state Structured P2P Networks Pastry Routing

91 Node Addition in Pastry New node: d46a1c d46a1c d471f1 d467c4 d462ba d4213f Route(d46a1c) d13da3 65a1fc Structured P2P Networks Pastry Node Addition

92 DHT Schemes for Structured P2P Networks Chord Pastry Tapestry Content Addressable Network (CAN) Structured P2P Networks DHT Schemes

93 What is Tapestry? A prototype of a decentralized, scalable, fault-tolerant, adaptive location and routing infrastructure (Zhao, Kubiatowicz, Joseph et al. U.C. Berkeley) Network layer of OceanStore global storage system Suffix-based hypercube routing Core system inspired by Plaxton Algorithm (Plaxton, Rajamaran, Richa (SPAA97)) Core API: publishobject(objectid, [serverid]) sendmsgtoobject(objectid) sendmsgtonode(nodeid) Structured P2P Networks Tapestry Overview

94 Tapestry Uses plaxton mesh data structure allows nodes to locate objects and route messages to them across an arbitrary-sized network while using a small, constant-sized routing map at each hop route map is called neighborhood map Ideas similar to pastry Structured P2P Networks Tapestry Overview

95 Plaxton (Suffix) Routing Example from to destination path resolution order: xxxx7 xxx67 xx567 x x represents wild cards All the numbers are in base-10 Structured P2P Networks Tapestry Plaxton Routing

96 Example Neighborhood Map Suffix routing Each entry points to next node in the neighborhood A neighbor is a connected node to the present node Level corresponds to the number of matching digits Neighbor Map For 5712 (Octal) x012 x112 x212 x312 x412 x512 x xx xx22 xx32 xx42 xx52 xx62 xx72 2 Routing Levels xxx0 xxx xxx3 xxx4 xxx5 xxx6 xxx7 1 Structured P2P Networks Tapestry Neighborhood Map

97 Another Suffix Routing Example Example: Octal digits, 2 18 namespace, Structured P2P Networks Tapestry Plaxton Routing

98 DHT Schemes for Structured P2P Networks Chord Pastry Tapestry Content Addressable Network (CAN) Structured P2P Networks DHT Schemes

99 Content Addressable Network (CAN) Associate to each node and item a unique id in an d-dimensional space Properties Routing table size O(d) Guarantee that a file is found in at most d*n 1/d steps, where n is the total number of nodes Structured P2P Networks CAN Overview

100 CAN Example: Two Dimensional Space 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) Node n1:(1, 2) first node that joins cover the entire space n Structured P2P Networks CAN Example

101 CAN Example: Two Dimensional Space Node n2:(4, 2) joins space is divided between n1 and n n1 n Structured P2P Networks CAN Example

102 CAN Example: Two Dimensional Space Node n2:(4, 2) joins space is divided between n1 and n n n1 n Structured P2P Networks CAN Example

103 CAN Example: Two Dimensional Space Nodes n4:(5, 5) and n5:(6,6) join n3 n4 n n1 n Structured P2P Networks CAN Example

104 CAN Example: Two Dimensional Space Nodes: n1:(1, 2); n2:(4,2); n3:(3, 5); n4:(5,5);n5:(6,6) Items: f1:(2,3); f2:(5,1); f3:(2,1); f4:(7,5); n3 n4 n5 f f1 n1 n2 f3 f Structured P2P Networks CAN Example

105 CAN Example: Two Dimensional Space Each item is stored by the node who owns its mapping in the space n3 n4 n5 f f1 n1 n2 f3 f Structured P2P Networks CAN Example

106 CAN: Query Example Each node knows its neighbors in the d-space Forward query to the neighbor that is closest to the query id Example: assume n1 queries f n3 n4 n5 f f1 n1 n2 f3 f Structured P2P Networks CAN Example