Download presentation
Presentation is loading. Please wait.
Published byErik McCoy Modified over 8 years ago
1
Cloud-based movie search web application with transaction service Group 14 Yuanfan Zhang Ji Zhang Zhuomeng Li
2
School of Electronic Engineering and Computer Science Outline Motivation Goals System design Micro-Scope & Macro-Scope Results and contribution 2
3
School of Electronic Engineering and Computer Science Motivation With the increasing demand for the capacity of the web servers, load balancing will be a crucial part in managing work load on each server to reduce traffic congestion and avoid break-down failure. In order to make the web servers to serve our Movie Online Search and Sale Website with high reliability and QoS satisfaction, we were inspired by a paper called ‘Dynamic Load Balancing on Web Server Systems’. It involves various of TTL algorithms and descriptions for multiple loading request approaches with throughput and availability improvement. Thus, we aim to select one appropriate load balancer from the paper to assist our original Movie E-commerce Web App in loading, scheduling and distributing requests to different servers with availability and performance enhancement.
4
School of Electronic Engineering and Computer Science Goals 1. Basically build a Movie Search & Sale Web App by use one instance in AWS EC2 cloud at first. 2. Survey on various of loading request approaches and file replication research, analyze them in current specific scenario, and select an appropriate one with simple but efficient consistency scheme. 3. Scale up web servers as multiple instances and incorporate with specific load balancer selected. 4. Evaluate by using Jmeter. And verified as performance improved.
5
School of Electronic Engineering and Computer Science System design (Micro-scope) User Case Diagram for Web Application 5
6
School of Electronic Engineering and Computer Science Implementation result 6
7
School of Electronic Engineering and Computer Science System design (Macro-scope) Architecture as a distributed system 7
8
School of Electronic Engineering and Computer Science Performance analysis 8 Single-instance CasesScaled-version Cases 1 Thread 10 Threads
9
School of Electronic Engineering and Computer Science Case 1: HTTP/ 1 Thread Case 2: HTTP/ 10 Threads Query Time Search Servlet Time JDBC Time 97 ms 140 ms 3.07 ms3.3 ms 2.04 ms1.37 ms Single-instance Cases
10
School of Electronic Engineering and Computer Science Case 1: HTTP/ 1 Thread Case 2: HTTP/ 10 Threads Query Time Search Servlet Time JDBC Time 146 ms135 ms 1.34 ms2.0 ms 0.49 ms0.82 ms Scaled-version Cases
11
School of Electronic Engineering and Computer Science 11 1 Thread Case 10 Threads Case
12
School of Electronic Engineering and Computer Science References Dynamic Load Balancing on Web-Server systems, By Valeria Cardellini, Philip S. Yu Web-server workload characterization: the search for invariants, By M.F. Artlitt Enhancing the Web infrastructure: from caching to replication, By M. Baetchsch Cluster-Based File Replication in Large-Scale Distributed Systems The google file system, By Sanjay Ghamawat, Shuntak Leung 12
Similar presentations
© 2025 SlidePlayer.com. Inc.
All rights reserved.