Presentation is loading. Please wait.

Presentation is loading. Please wait.

Meng Cao, Xiangqing Sun, Ziyue Chen May 28th, 2014

Similar presentations


Presentation on theme: "Meng Cao, Xiangqing Sun, Ziyue Chen May 28th, 2014"— Presentation transcript:

1 Meng Cao, Xiangqing Sun, Ziyue Chen May 28th, 2014
Wireless Network that Forms a Simple Cluster among Computers and Mobile Devices Meng Cao, Xiangqing Sun, Ziyue Chen May 28th, 2014

2 Project Overview Explore the feasibility of utilizing the computation potential of personal mobile devices (laptops, tablets, or mobile phones, etc) to form a local Hadoop cluster. All devices will be connected through WiFi Ad-hoc network with one master node and multiple slave nodes. Master node’s responsibilities: Split a given task into sub-tasks; Assign sub-tasks to slave nodes; Merge returned partial results from slave nodes Monitor network status, slave node connectivity Slave node’s responsibilities: Compute the results of a given sub-task Transmit the partial result back to master slave

3 Motivation Big Data Expensive, Enterprise Level Machines
Unavailable to public (like us!) Growing Performance of Smartphones CPU storage network OS Available Open-Source Platforms Apache Hadoop MapReduce, Hadoop Distributed File System (HDFS) Android SDK Free, Integrated IDE in Eclipse, Emulators

4 Motivation

5 Background-AD HOC What is Ad Hoc?
decentralized wireless network & no pre existing infrastructure How does it work? routing by forwarding data for other nodes decision made dynamically & flooding The Wireless Routing Protocol(WRP) table-based protocol each node maintains 4 tables: 1. distance table 2. routing table 3. link-cost table 4. message-retransmission list table communicate through update msgs

6 Background Hadoop Framework

7 Project Setup WiFi Ad-hoc network setup
Enable WiFi Direct on both Debian laptops and Android phones Hadoop setup on Linux machines Hadoop single node setup Hadoop cluster setup among Debian laptops in LAN Simple task programming and testing Configure Hadoop cluster on Ad-hoc networks Porting Hadoop framework onto Android devices (More details on the next slide) Testing & Simulation Few nodes running test (One laptop as master node, 2-3 Android phones as slave node) More nodes simulation through Android emulator on PC

8 Port Hadoop to Android Modify & re-compile Hadoop source code for Android compatibility Modify & re-compile Java libraries required by Hadoop DataNode & TaskTracker processes are replaced with service processes Performance tuning Memory allocation CPU utilization

9 Thank you!


Download ppt "Meng Cao, Xiangqing Sun, Ziyue Chen May 28th, 2014"

Similar presentations


Ads by Google