2. Peer-to-Peer By Rui Zhang, Chen Teng, Li Dong, Quanshuan He & Yongzheng Zhang

3. Overview What is peer-to-peer Application Advantages and Disadvantages Case Study(Gnutella) Conclusion References

4. What is peer-to-peer Why P2P? Circumstance Concept Landscape

5. Why P2P? Growing application demand Good Suitability Update immediately

6. Circumstance Napster – focus awareness One of important Internet service foundations Part of other architectures application process Scale of peer-to-peer protocol

7. Circumstance(cond.) Statistic number Home users: 79.5 million Work users: 1.2 million Simultaneous users: 640,000 Downloads in September alone: 1.4 billion

8. What is Peer-to-Peer? Traditional Peer-to-Peer "A type of network in which each workstation has equivalent capabilities and responsibilities. This differs from client/server architectures, in which some computers are dedicated to serving the others. "

9. What is Peer-to-Peer?(cond.) P2P Architecture Today key characteristics: interfaces running outside of a web browser both clients and servers easy to use and well-integrated content creation or functionality addition support connections provide something new! "cross-network" protocols support(SOAP or XML-RPC)

10. Landscape P2P Distributed Computing P2P Affinity Communities Peertailing

11. Key features of P2P application Discovering other peers Querying peers for content Sharing content with other peers

12. Different types of P2P application Pure P2P P2P with a Simple Discovery Server P2P with a Discovery and Lookup Server

13. Pure P2P No central server How to discover peers Uses information from local configuration scheme Employs network broadcasting and discovery techniques Limits the application’s reach

14. Pure P2P(cont.) Peer 2 Peer 3 Peer 1 (1) Content query (2) File transfer

15. Pure P2P Peer 2 Peer 3 Peer 1 (1) Content query (2)Content query (3) Response (4) Response (5) Connect and File transfer

16. P2P with a Simple Discovery Server Notifies central server of its existence at startup time Uses central server to download a list of other peers Goes through the list and contacts each peer individually with its request

17. P2P with a Simple Discovery Server Peer 1 Peer 2 Peer 3 Server (2) Peer list (1) Log in (3) Content query (4) File transfer

18. P2P with a Discovery and Lookup Server Server includes both discovery and content lookup services The peer application registers with a discovery server and uploads a list of its contents at regular intervals Queries central server for particular content Reduces the number of queries

19. P2P with a Simple Discovery and Lookup Server Peer 1 Peer 2 Peer 3 Server (2) List of peers which have the requested content (1) Tell server which content it wants (3) Content query (4) File transfer

20. When Using Napster  You need:  A copy of the Napster utility installed  A directory on your computer that has been shared  Some type of Internet connection

21. When Using Napster(cont.) The provider of the song needs: A copy of the Napster utility installed A shared directory on their computer Some type of Internet connection that is currently on A copy of the song you are looking for in the designated shared directory

22. The Napster Network

23. P2P advantages & disadvantages

24. P2P advantages Low cost Sharing individual resources ---- data resources (Napster) ---- hardware (SETI) Administration Highly fault-tolerant Real time updating (online auction)

25. P2P disadvantages The limited access number Availability Hard to predict the consequences of failure Bandwidth consumption Security problem

26. Security Problem Why client/server is more secure --------- centralized resource --------- centralized administration --------- system integrity

27. Why P2P is less secure --------- non-specialist users --------- vendors --------- authentication information --------- disclosure of IP and MAC addresses --------- virus distribution Security Problem (cont.)

28. Security problem ---- possible solutions Limit and restrict access number ----- validate certificate ----- obtain certificate ----- caching data (FreeNet)

29. Case Study Gnutella

30. Concepts 1. Introduction for Gnutella 2. Gnutella & Firewalls 3. Security Considerations for Gnutella Users 4. Gnutella Protocol Information 5. Limitions and Risks for Gnutella

31. Gnutella basics (1)An open, decentralized, peer-to-peer search system. It is a name for a technology. (2)The Gnutella protocol and original servent ("Gnutella 0.56") were conceived and developed by Justin Frankel and Tom Pepper at Nullsoft in March, 2000. (3) Each piece of Gnutella software servent (SERVer+cliENT) is both a server and a client in one.

32. Gnutella 1. Gnutella Is File sharing. 2. Gnutella Is Anonymous. 3. Gnutella Is The Game : Telephone. 4. Gnutella Is Designed to Survive Nuclear War. 5. Gnutella Can Withstand A Band of Hungry Lawyers.

34. How Gnutella retrieves information

35. Gnutella & Firewalls 1. With a firewall, there are some problems for Gnutella when making a request for a file. 2. To compensate for this, Gnutella's designers came up with the "push request".

36. How Gnutella handles firewalls

37. Internet Security Considerations 1. IP Address Advertising 2. Connection Acceptance

38. IP Address Advertising 1. Peering networks dynamically collect, distribute, and broadcast the IP addresses of their active peers. 2. Malicious hackers now use special "IP Address Harvesters" to collect the Internet addresses of active, online, peering clients and servers,then target them by their IP addresses for direct attack.

39. Connection Acceptance 1. The typical personal computer never need to accept unknown connections. 2. Users of peering services such as Gnutella do accept connections from other unknown machines and are therefore temporarily acting as Internet servers which are similarly vulnerable to direct attacks.

40. What Can You Do? 1. Take responsibility and get yourself informed! 2. Get your Shields UP! 3. Add a free Firewall! 4. Ignore the IBR(Internet Background Radiation)! 5. Tell Your Friends!

41. Gnutella Protocol Information

42. General Description Works by “Viral Propagation” Inordinate amounts of traffic In reality, it isn’t so bad! (Horizon 10000) Uses GUID to identify each message Each servent maintains a short memory of GUIDs it has seen http

43. Connecting to a Servent Connect to other gnutella servents: GNUTELLA CONNECT/0.4\n\n The accepting servent responds: GNUTELLA OK\n\n After that, it's all data.

44. Gnutella Messages Data passed on the Gnutella network are called "messages” (Header+Payload): 1. PING request 2. PONG reply 3. Query (Search Request) 4. Query Hit (Search Reply) 5. PUSH request

45. Header Format BytesSummaryDescription 0-15Message Identifier GUID, used to identify each particular message 16 Payload Descriptor (Function Identifier) ValueFunction 0x00Ping 0x01Pong 0x40Push Request 0x80Query 0x81Query Hit 17TTLTime To Live (hops left before dropped) 18HopsNumber of hops this message has taken 19-22Payload Length The length of the data which follows the header

46. Ping (function 0x00) No payload Servent sends/forwards PING message to all connected servents

47. Pong (function 0x01) Payload Routing Instruction: Servent sends/forwards Pong message back along the path its Ping came from. BytesSummaryDescription 0-1Port numberPort number of responding host 2-5IP addressIPv4 address of responding host 6-9# of filesNumber of total files shared 10-13# of kilobytesSize of total files shared

48. Query (function 0x80) Payload Routing Instruction Servent sends/forwards Query message to all connected servents. BytesSummaryDescription 0-1Minimum Speed The minimum speed, in kilobytes/sec, of responding hosts 2+Search Criteria Search keywords or other criteria. NULL terminated.

49. Query Hit (function 0x81) Payload BytesSummaryDescription 0 # of hits (N)# of hits in the result set following this header 1-2 PortIPv4 port number of responding host 3-6 IP addressIPv4 address of responding host 7-10 SpeedSpeed of responding host, in kilobits/s 11+ Result Set (N of these) last 16 Servent IdentifierGUID of responding host, used in PUSH BytesSummaryDescription 0-3IndexIndex number of file 4-7SizeSize of file in bytes 8+File NameTerminate:double NULL

50. Query Hit (cont.d) Routing Instruction Servent sends/forwards QueryHit message back along the path its Query came from.

51. Push Request (function 0x40) Payload Routing Instruction Used when trying to download a file from the servent behind a firewall Push messages is sent along the path on which the query hit was delivered. BytesSummaryDescription 0-15Servent IdentifierGUID of the servent which should push 16-19IndexIndex number of file (given in query hit) 20-23IP addressIPv4 address of servent to push to 24-25Port numberIPv4 port number of servent to push to

52. Routing Examples A BC DD CB A

53. Imagine yourself as node 1. You have direct (physical socket) connections to nodes 2, 3, 4, and 5. You have reachable hosts at nodes 6 through 13. 1. You get a Ping from 2 with GUID of x. 2. Lookup in your routing table [message, socket] 3. Not there? Save [message x, socket 2] in the routing table. 4. Respond with a Pong (GUID x) to node 2. 5. Forward this Ping to nodes 3, 4, and 5 (not 2!!). 6. Node 3 will respond with Pong (GUID x) to you. 7. Record [message x, socket 3] in routing table, then fine the entry [message x, socket 2], so forward this Pong to node 2. 8. Do the same thing with responses from 4 and 5. 9. Since node 3 through 5 will also pass the Ping on to 8 thru 13, you'll also get a Pong from them too. 10. Node 3 is connected to 10 who is connected to 4 and 4 is connected to you! Node 4 will also send a Pong message along the path 4  10  9  3  1. You lookup in your routing table and find [message x, socket 4] is already there! You drop the message, and do not forward to anyone!

54. Downloading File The servent requests the file using HTTP: GET /get/1234/blue.mp3 HTTP/1.0\r\n Connection: Keep-Alive\r\n User-Agent: Gnutella\r\n Range: bytes=0-\r\n \r\n The servent will respond with normal HTTP headers, e.g.: HTTP 200 OK\r\n servent: Gnutella\r\n Content-type:application/binary\r\n Content-length: 1624\r\n \r\n Supports the range parameter to resume partial downloads

55. Topology Summary Gnutella network has no hierarchy, i.e. every servent is equal. Some servents contribute more than others. Gnutella network is not a tree and it is cyclic. Gnutella is barely HTTP.

56. Limitations and Risks Problem in scaling (not a tree) TTL imposes a horizon (10000) on each user Hackers misuses Gnutella for other reasons Difficulty in authenticating the source of the data returned

57. Conclusion Peer to peer is now being recognized as the computing paradigm of the future.

58. References http://www.peer-to-peerwg.org http://www.gnutellanet.com http://www.gnutellanews.com http://gnutella.wego.com http://www.limewire.com/glossary.htm