• Tidak ada hasil yang ditemukan

Chapter 1 Introduction

2.2 Summary

CHAPTER 2� LITERATURE SURVEY 36

decided by the entries in the proximity list. The closest node to the target key in key space will be found out as next hop from local entries of both finger list and proximity list of the current node. Then the entry with that identifier will be chosen as the next hop.

6. HPS [GEBF+03]: The general framework for hierarchical p2p system has disjoint clusters. In this system lookup messages are first routed to the destination cluster using a inter-cluster overlay, and then routed to the destination peer through a intra-cluster overlay. Each cluster has one or more superpeers which are selected from certain criteria, e.g. reliability and connectivity. Each normal peer chooses to join one cluster based on the specific requirements of the application. Superpeers form the upper- level overlay. Each cluster can use autonomous intra-cluster overlay lookup service.

7. HIERAS [XMH03]: It is a multi-layer DHT that organizes the partici- pating peers into multiple overlay networks (rings) at the different layers.

Each of these rings contains a subset of all system peers. Several lower level p2p layers are created, according to the topological proximity metric, besides the upmost level of p2p layer which contains all the participating peers. A routing procedure is executed at the lowest level firstly, and then moves up to a higher level and eventually reaches the upmost level layer.

So, the lower the layer of a ring, the smaller is the average link latency between two peers participating in that ring. Thus a lower total routing latency can be achieved.

CHAPTER 2� LITERATURE SURVEY 37

RChord DRChord Chord MLChord PChord HPS HIERAS

Architec- ture

Overlay is cate- gorized into three dif- ferent types

Based on sim- ple chord but the chord ring is di- vided into two differ- ent ring structure

Proposes two tier architecture

Multilevel chord protocol

Uses proxim- ity list to evaluate topology of the un- derlying network

Different peers are orga- nized into some disjoint clusters

Multilayer DHT protocol that organize peers into mul- tiple overlay networks Lookup

Protocol

Different routing protocol is carried out on overlays

Routing is done on two differ- ent rings based on their existence

First routing is done at the up- per level formed by the stbale nodes, then to the next lower level

Routing is carried out based on the rout- ing table on each node in a layer

Next hop is chosen from the prox- imity list for efficient routing

Each cluster uses au- tonomous intra- cluster overlay lookup service

Starts at the lower level first, then moves upto the higher level

Table 2.3: Comparison of Hierarchical P2P protocols

CHAPTER 2� LITERATURE SURVEY 38

found that p2p overlays are increasingly becoming popular choice for most of the Internet based applications. Among these the hierarchical overlays are the most recent trend in p2p area. The heterogeneity of nodes leads one to treat the nodes of p2p systems in many different ways. it is observed from the survey that some of the main challenges that p2p systems should face are - scalability, performance, adaptability, load balancing, resilience, and providing security and incentive mechanism.

Chapter 3

SHP: A Structured Hierarchical Overlay Protocol

The structured overlays use a specific routing geometry and also support key- based routing in a way that object identifiers are mapped to the peer identifier address space and an object request is routed to the nearest peer in the peer address space. Peer-to-peer systems using key-based routing are also known as distributed object location and routing (DOLR) systems. The distributed hash table (DHT) is a specific type of DOLR.

The structured overlays are designed with many dimensions. The design of structured overlay systems are based on maximum number of hops taken by a request on an overlay ofn peers, organization of the peer address space, next-hop decision criteria, geometry of the overlay, overlay maintenance and locality and topology awareness. The identifier space is large in a p2p system. The address space may be flat or hierarchical. Hierarchical overlay consists of n-tier overlays whereby the nodes are organized into disjoint groups. The overlay routing to the target group is done through an intergroup overlay; then intra-group overlay is used to route to the target node. The hierarchical overlay architecture of the Internet offers several important advantages over the flat DHT-based p2p overlay

CHAPTER 3� SHP: A STR� HIERARCHICAL OVERLAY PROTOCOL 40

such as reduces the average number of peer hops in a lookup query and minimizes the query latency. The hierarchical overlay facilitates large-scale deployment by providing administrative autonomy and transparency while enabling each partic- ipating group to choose its own overlay protocol. Intra group overlay routing is totally transparent to the higher-level hierarchy. If there were any changes to the intra-group routing and lookup algorithms, the change is transparent to other groups and higher-level hierarchy.

In a p2p system large number of autonomous systems participate and form self organizing networks, which are based on the overlay networks i.e. over the top of conventional IP with no centralized structure. The distributed object lo- cation service is used to establish communication among nodes, which in turn requires efficient routing for better performance of the system as a whole. The participants of a p2p systems are typically individuals and organizations without an out-of-band trust relationship and systems are heterogeneous in nature.

The aim of this chapter is to develop aStructured Hierarchical overlay Proto- col (SHP) to improve the overall performance of the system. The overlay that is proposed should be able to handle both point and range queries and also should be capable of handling transient node population in the system with inherent hierarchical property. One of the major applications of p2p systems is content distribution. Millions of nodes may join a p2p network to share resources spread over the world. It is detrimental to the performance of the system if all the nodes are considered identical. Nodes in a network may differ in terms of available bandwidth, processing power, stability, privacy & security, reliability, amount of data shared and storage capacity.

The nodes in the system are classified as given in section 3.1 and the system architecture is discussed the system architecture in section 3.2. The process of

CHAPTER 3� SHP: A STR� HIERARCHICAL OVERLAY PROTOCOL 41

joining and leaving of nodes and query processing are discussed in sections 3.3 and 3.4, respectively. A load balancing mechanism is proposed in section 3.5.

The storage requirement of the system is discussed in section 3.6. The section 3.7 provides experimental results and related works for this chapter are presented in section 3.8. Finally, the summarization of this chapter is done in section 3.9.