Transcript How p2p
P2P设计思想及其 在存储和共享上的应用 安全组: 吕建明 2004-3 outline What is p2p Why p2p How p2p P2p storage and sharing –my work Road map What is p2p 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. ―――http://www.webopedia.com/ What is p2p Peer-to-peer is a communications model in which each party has the same capabilities and either party can initiate a communication session. Other models with which it might be contrasted include the client/server model and the master/slave model。 On the Internet, peer-to-peer (referred to as P2P) is a type of transient Internet network that allows a group of computer users with the same networking program to connect with each other and directly access files from one another's hard drives ―――http://searchnetworking.techtarget.com/ powered by whatis.com What is p2p Feature of p2p communication model Peer interconnection interoperation outline What is p2p Why p2p How p2p P2p storage and sharing –my work Roadmap Why p2p 集中式 •单点错误 •维护成本高 •扩展耗费大 •性能瓶颈 Why p2p p2p •Self-organized •mobile computing •self-cooperation •Scalability •fault-tolerance •cost effective Why p2p p2p is a communication model , u can use it every where to remove the centralized point : web service / storage system / grid computing/ office ware / instant messaging / peer is different . (from the physical layer to the logical layer) . There is no common p2p platform can suit every use case. Why p2p while the typical communication model is is a type of transient Internet network that allows a group of computer users with the same networking program to connect with each other and directly access files from one another‘s hard drives the internet (like Kazza/Napster/pp点点通/edonkey) outline What is p2p Why p2p How p2p P2p storage and sharing – my work Roadmap How p2p - JXTA There are a lot of different scenarios to deploy p2p. application JXTA (sun p2p platform) display some common scenarios and brings up some common problems . www.jxta.org 80 projects have been built upon JXTA How p2p – JXTA (2) How p2p – JXTA (3) JXTA 定义了p2p领域中的一些通用的问题： peer addressing , Peer group , resource location , security , communication model , security , peer monitoring , etc. 并给出了参考的协议实现。JXTA for J2SE , etc 但不是万能钥匙。具体的p2p系统有不同的需求。如： 特殊的安全需求、特殊的路由定位需求、特殊的效率 需求、特殊的通信模式的需求 不存在统一的平台 How p2p – Common steps Find the centralized point and remove them direction server / schedule server / monitor server / indexing server make sure what is the peer and how many types of peer. How p2p – Common steps how peers communicate (communication model) What is communication message How does peer discover each other and resource Security issues … … Build the p2p middle ware How p2p – communication model Unstructured p2p networks Structured p2p networks Loose structured p2p networks Unstructured networks 网络形成：节点采用随机的方法或采用启发策略加入 网络，网络拓扑随着节点的变迁和网络通信的进行而 发生演变。 这一类的系统包括Gnutella、FastTrack、kazaa、 Limwire、Usenet 、freenet、Planetp等等。 路由特征：带启发的广播查找的过程，启发的方法包 括span ring、degree based、supper nodes 、 routing index、LSI method、bloom filter 等等。 往往引入集中节点来提高效率。（hybrid p2p) Unstructured networks – pure p2p Unstructured networks – hybrid p2p Unstructured networks – hybrid p2p(2) Structured networks CAN拓扑－n维笛卡儿空间 CHORD拓扑－环形结构 Structured networks(2) ☆ 网络拓扑特点：每个节点都有固定的编址，整个网络 具有相对稳定而紧致的拓扑结构， ☆ 数据存储特点：DHT（Distributed Hash Table） ☆ 网络路由：通过O(lg N)跳就可以定位目标节点。（N 是总节点数） ☆ 这一类的系统包括CAN、Tapestry、Pastry、kademlia、 Chord等 Structured networks - DHT key7 key8 key9 data7 data8 data9 d c f b key10 key11 data10 data11 key12 key13 key14 data12 data13 data14 key4 key5 key6 a e key0 Peer addr e f data4 data5 data6 key1 key2 key3 data0 data1 data2 data3 key 0 1 d 4 c 7 b 10 a 12 Structured networks – DHT(2) 快速数据定位方法： DATA KEY Route (KEY) Reach the node whose address is closest to KEY Loose Structured networks key1 key2 key3 data1 data2 data3 Loose Structured networks(2) ☆ 网络拓扑特点：unstructured networks ☆ 数据存储特点：DHT ☆ 例子：JXTA 的集中节点之间的通信 Loose structured outline What is p2p Why p2p How p2p P2p storage and sharing - my work Roadmap P2p storage and sharing File sharing on internet is quite popular. Search and download movie /mp3 from others Unstructured hybrid p2pnetworks can do with this . Centralized servers are used to promote the performance. A lot of clients. Gnutella, FastTrack, kazaa, Limwire, edonkey, pp点点通 etc. P2p storage and sharing But unstructured p2p does not fit a more reliable and high performance system design: auto content indexing & content searching reliable ocean storage reliable p2p computing environment Use case study How to organize a lot of machine to let them compose a system ,which is auto content distributing / indexing /searching high scalability , auto p2p backup , self-organized resource discover , self-organized resource scheduling , self-organized monitoring, … … Use case study (2) Benefit – Make full use of computing and storage resource Reliable Convenient for maintenance Much cheaper What we need to do ? Use Structured p2p networks for communication model , because : 相对稳定而紧致的拓扑结构适宜做稳定的服务 DHT数据存储，提高数据检索速度 高效网络路由(o(logn))适合做自组织的资源调度、移 动计算 What kind of structured p2p Tapestry (UCB) ? CAN (UCB) ? Pastry (Rice) ? Chord (MIT)? Or all of them ? Difference between them A Peer how to find another peer in o(logn) How to form the local routing table of peer ? Tapestry /pastry routing table Tapestry /pastry Node 5230 find node 42AD CAN Neighbor :节点地址除了在某 一个位上数字相邻，其它位 的数字都相同： 节点0011的邻居有： 1011、0111、0001、0010 Routing table : IP Address of neighbor set Routing : according the logical n-cartesian coordinate space We will use CAN 地址向量相近的节点互为邻居。 1011、0111、0001、0010 （chord/tapestry/pastry是线性相近 1000 、1001、 1002、1003是邻居） 从而在DHT中，key向量相似的data分布在相邻的节点 上，有利于做内容检索。 格状拓扑，稳定灵活（比tapestry/pastry稳定，比 CHORD灵活），适合组织调度 CAN is perfect? routing according to the logical space only Gap between logical and physical make routing ineffective Make no use of local physical information Improve CAN Extent Routing table entry to: (peer address , ip address , pingtime) (like tapestry) Extent routing table with local neighbor set ,which recording the closest peer on physical networks layer. (like Pastry) Pastry routing table What we will make in the end A middle ware providing API of: Grid topology connection Routing content indexing fast Resource searching auto p2p backup auto resource schedule middle-ware use case entrance Distributed according to the key index middle-ware use case auto content searching middle-ware use case P2p auto backup middle-ware use case auto finding the idle machine middle-ware use case fault- tolerance What it will look like? Like Grid computing ! But more self-organization But more p2p character outline What is p2p Why p2p How p2p P2p scenarios P2p storage and sharing –my work Road map Roadmap Build the basic routing module . (modify the routing core of tapestry open source project) Build the content-searching module Build the auto backup module Build the resource schedule module Share it … Thank u for any further advice !