The Advantage Of Performance In Service Discovery

Published Date: 02 Nov 2017

2. Literature Review 3

3. Project Scope and Objective 5

3.1Objective 5

3.2 Scope 6

4. Method and Technology Involve 7

4.1 Method 7

4.2 Technology Involve 7

5. References 9

1. Project Backgroud:

P2P technology are start many year ago. From the previous Napster (drscholl, 2001), Gnutella and Freenet until now we have BitTorrent (BitTorrent, 2013) which is one of the most common P2P file sharing application. In the pass, Peer-to-Peer (P2P) are commonly use on file sharing application, but nowadays, many different P2P application allow users to share the computation power and bandwidth and etc. Moreover, some system now adopted usage of P2P application to improve the scalability of the system. One of the example is Skype which the bandwidth are shared between users. The widely used of P2P in different application is because the P2P architecture enable the scalability, performance and efficiency to the system.

In the P2P application, one of the most important criteria to evaluate the efficiency of application is the effectiveness of discovery method to get the resources in the P2P network. There are many type discovery method (Ed So, 2002) used in different P2P application. For example:one the method is Centralized indexes and repositories. Napter uses this method to keep all the index of all peers and their resources on a main server. When a client request a service, the query will send to server, then server will lookup for the index and after the query solved, the the server will send a message to client about where can get the resources. Centralized indexes have the advantage of performance in service discovery. However, it is very expensive to have a server can handle large network of peers. Second example of service discovery method is flooding broadcast of queries which is used by Gnutella. When a peer in network make a query, then the query will broacast to all the neighbor peers. If the query dint not solve, the query will continue broacast until found the resources. After that, a message will send back to the original sender, and then establish the connection between sender and resources. This method has the advantage of flexibility in processing the query, however, the broadcast of the query will cause the network congestion. Besides that, there are many other service discovery method which will carring its advantage and drawback.

In the other hand, P2P system architecture also integrate with cloud computing to solve the issue in the cloud computing. P2P architecture can solve the the problem of scalability, single point of failure, network congestion in the cloud computing.

The rest of this report is organized as follows. Section 2 is literature review of unstructure P2P and structure P2P and section 3 is the project scope and objective. In section 4 is describes the method and technology involve in this project.

2. Literature Review

In general, P2P can be categorized into unstructured and structure P2P. Some of the researchers have research on the using the unstructured P2P approach to solve the scalability in cloud. Besides, nodes can join or leave the network anytime as they wish, it cause the unstructured P2P does not have constant network structure, so unstructured P2P require less maintenance. However, unstructured P2P also have several issues. Firstly, when unstructured P2P nodes need to search the particular service or resource in the network, it involves blind search and informed search, because nodes in the P2P network have no information about other nodes. However, the informed search will keep some data location information. Secondly, when nodes in unstructured P2P network preform searching for certain resources in the network, it will uses message flooding by broadcasting message to all directly connected in the network. This will cause a problem when there are too many requests in the unstructured P2P network, the number of network message would be become too large and it will cause the traffic congestion. The following is a few example of proposed solution using unstructured P2P:

Zhou (Zhou, et al., 2011) proposed solve the scalable mechanism for service discovery by utilizing semantic Web technologies to describe Cloud services in a semantic manner. Due to the multi-dimensional of the semantic description of Cloud services and the user request for the services, they apply the kids-tree to establish a data space to create indices for all the services and developed an efficient hybrid P2P search scheme. This paper combination of one-hop replication, semantic-aware routing, shallow flooding, topology re-organization and supernodes. The weakness of this proposed solution is it cannot totally prevent the message flooding which will cause network traffic.

On the other hand, structure P2P system also research by many researcher to overcome the problem mention above. Structure P2P systems sometime also referred as Distributed Hash Tables (DHTs). Structure P2P systems are scalable network infrastructure that support large-scale distributed systems. Moreover, Structured P2P system also provide better search which it can determine which node should go when send the service request. Because the neighborhood links are better defined in structure P2P network. The following example is the approach using Structure P2P:

Chord4S (Qiang He, 2008) is a Structure P2P which a chord based decentralized service discovery. This paper is use chord to utilize the data distribution and lookup capabilities to distribute and discover services in a decentralized manner. Chord4S improve the data availability by distributing service description of functionally-equivalent services to different successor nodes that are organized into a virtual segment in the chord circle. Chord is selected in this paper is because it is flexibility and scalability, this can be considered suitable in large-scale network. Besides, Chord also have advantage in node organization, data distribution and query routing to improve efficiency of service discovery. Furthermore, the chord routing protocol is extended to support efficient discovery of multiple services with single request. Weakness of this proposed solution is the locality do not consider in this solution. So, when the service consumers request a service, it service provider may just nearer in geographically buy it route to a service provider which is located far away.

3. Project Scope and Objective

This project is extended from existing proposed solution by Soung-Yue Liew (Soung-Yue Liew, 2012) which is a new service discovery approach based on Pastry. Pastry is a structure peer-to-peer network model. They proposed a directory node as platforms which is use to advertise service with first m bits of the service ID as a key (prefix). Then consumers need to discover the service in the network by using the key to find the directory node and then get the service provide node. In the proposed P2P-Cloud framework, all virtual machines running in a cloud environment on heterogeneous hosts are considered as peer nodes. Each node has its own resources to share and request resources from other node. Moreover, each peer nodes ID are generate by hash function to have unique NodeID and ServiceID. Pastry is used to maintain locality while performing routing. Moreover, to reduce the complexity, they propose a fixed-length prefix, the first m bits of the service ID to be used to service advertise and service discovery.

3.1Objective

The main objective of this project is use FreePastry as a platform to run the simulation to evaluate the performance of the proposed scheme. The evaluation criteria consist of service availability and the corresponding success rate in advertising and retrieving the services from the directory nodes. Besides that, the routing performance in term of hop count and latency between nodes are evaluated in the event of successful service discovery and service advertisement.

To evaluate the performance of the scheme, different number prefix length for key (m bits) are set. This will used to test which prefix length have the highest percentage of advertisement success rate and service discovery success rate. Besides that, the routing performance are evaluated in different prefix key length.

To simulate how the number of peer nodes in the network (network size) affecting the success rate of advertisement and service discovery with the different m bits.

To simulate how the base used in the routing table, B affecting the performance of the service advertisement and discovery. The number of base used, will affecting the routing table size in each nodes.

3.2 Scope

This project aims to develop a simulation modules based on FreePastry. The module is used to simulate the proposed scheme and evaluate the performance in term of success rate of service discovery and advertisement discovery, and routing performance such as hop counts and latency. After that, study on the result of simulation to make further improvement of the algorithm of the proposed scheme.

4. Method and Technology Involve

4.1 Method

This section will mainly discuss what the method used and technology involve in this project to achieve project’s objective. This project scope are mainly conduct the simulation on the proposed scheme. Experimental research method will be used to achieve the objective. In order to start simulate the performance of the proposed scheme, first I need to set up the platform before running the simulation. In this project, FreePastry is used to run the simulation. There are a few step to set up the FreePastry simulation. First, a Virtual machines (VM) install on window. Then create a virtual machines and "OpenNebula 3.8 VirtualBox Sandbox" is use as virtual hard drive, then install the Ubuntu Linux OS into the virtual machine. Lastly, the FreePastry simulation will done on the created virtual machine.

The simulation will try with the different parameter, so see how the parameter affect the result of the simulation. These parameter involve is prefix length use for key (m bits), the size of the network (number of peer nodes inside the network) and base (B).These parameter will test through the experiment, and see how the parameter affect the success rate of the advertisement discovery and service discovery. The network performance in term of routing hop and latency also record to evaluate the performance. Lastly, the result obtain from the experiment will further study and used to make improvement of the algorithm in service advertisement and service discovery.

4.2 Technology Involve

This section will briefly describe the technology used in this project. First software use in this project is FreePastry (Hoye,J, 2001), it is an open-source implementation of Pastry P2P structured overlay network. This software is intended as a tool that allows researcher to evaluate Pastry. It is a platform used to perform further research and development of application in P2P substrates. And this software is written in Java programming language. FreePastry requires a Java runtime, version 1.5.0+ and developed using Sun’s SDK, version 1.5.0+. The version of FreePastry use in this project is version 2.1.

Second, this project simulation is run in Ubuntu Linux environment. The latest version of Ubuntu OS which is Ubuntu 12.10 is selected in this project. Ubuntu Linux OS is selected in this project is because Ubuntu open source operating system, it have more flexibility. Moreover, Ubuntu also more stable, high performance and more secure.