AN ANALYSIS OF USING WEB APPLICATION AS DISTRIBUTED COMPUTING PLATFORM Parin Chiamananthapong, Krerk Piromsopa Chulalongkorn University Bangkok, Thailand Parin.Ch@Student.chula.ac.th, Krerk.P@chula.ac.th ABSTRACT With the emerging in HTML5 and JavaScript, gap between web application and legacy desktop applications have been reduced; making this platform a suitable choice for distributed computing. In this paper, we present a survey of latest web technologies with the focus on performance improvement for distributed computing platform and use web browser as frontend client agent. We present an analysis of tools and compare to existing platform. Finally, performance evaluation has been performed. Several JavaScript features are gathered to benchmark the performance against that of the native compiled applications. The results show that performance achieved in web application is comparable to that of native applications. This platform eliminates the need to install additional software. It lowers the barrier of entrance. Clients can join by just simply opening a specific web page. In addition, it also gets the web application advantages including secure sandbox environment, instant deployment with platform independence and large user base. KEY WORDS Distributed Computing Systems, Web Technologies, Desktop Grid, Parallel Computing Systems 1. Introduction With the emerging in HTML5 and JavaScript, we want to prove that web technologies have been developed to the point that making it suitable for high performance computing. Desktop grid (aka. volunteer computing) is another powerful architecture. It is similar to traditional grid systems. However, the participants are internet-connected home desktops. The most successful project is SETI@home[1].The SETI@home project uses BOINC[2] framework. BOINC is designed to grab the large amount of spare processing power from massive number of internet connected computers around the world. However, it requires that each client must install dedicated software, which can raise some compatibility and security issues. Although all problems mentioned above can be solved by using virtual machine based desktop grid like Entropia[3]. It still needs a client to install software. This can discourage some participant. In order to lower the barrier of entrance for participants, we use web browser, the most usersโ€™ familiar platform that every computer or even a mobile device has it. With this approach, we can get the profit of the web application advantages, which are instant deployment, maintenance, and sandboxed environment. The participant does not need to install software or to perform additional work to participate. The only thing that a client needs to do is simply browsing a specific web page. In addition, participants can terminate the fail process or leave the computation anytime by closing the web browser without worrying about the damage to the system. With the portability of web applications and performance that comparable to native application, web applications have been used in other fields. For example, PhoneGap[25] use HTML5 for building mobile applications, mobile cloud computing application using HTML5[26], computing system with cloud-based browser[27] and web search using browser-based volunteer computing[28]. This paper makes the following contributions: ๎€ We present a survey of tools in a web application that enables implementing a distributed computing platform. Those tools are mainly designed for improving performance. We analyze and compare their characteristics. ๎€ We present an analysis of using a web application as distributed computing platform by comparing to existing platforms. Using web for computing has several advantages. However, there are some limitations and trade-off. We compare and show issues between web system and legacy system. ๎€ We perform two benchmarks and discuss their performance. The first benchmark is designed to compare the JavaScript performance to other computer programming languages. The other benchmark is for comparing the performance of JavaScript tools. We provide the analysis of each tool. The rest of paper is organized as follow. Section 2 is the related works. Section 3 is an introduction to JavaScript. Section 4 discusses tools for distributed computing in JavaScript and compares of web platform characteristics against traditional desktop grid. The performance evaluation and discussion are in section 5 Proceedings of the IASTED International Conference 2013) 1 - 1 , 2013 , Advan es c in Computer Science (ACS April 0 2 Phuket Thailand DOI: 10.2316/P.2013.801-080 434