130 likes | 241 Views
CS492: Special Topics on Distributed Algorithms and Systems Fall 2008. Lab 3: Final Term Project. Final Term Project. Goal In-depth study of distributed algorithms and systems
E N D
CS492: Special Topics onDistributed Algorithms and SystemsFall 2008 Lab 3: Final Term Project
Final Term Project • Goal • In-depth study of distributed algorithms and systems • Each team chooses a topic of interest related to distributed systems or algorithms or one from example projects • Each team will submit a term paper which includes: • Introduction and Motivation • Problem Definition • Related Work • Your Solution and Evaluation • Conclusions • Each team will have two presentations • First presentation should include: motivation, problem definition, and related work • Final Presentation at the end of semester
Evaluation Analysis Simulation Emulation Implementation
Project Ideas Storage Management PageRank in Large Dataset PageRankin Dynamic Dataset Paxos Project
1. Storage Management Many students faced with error caused by not enough disk storage while working on Lab 2. Total amount of storage of 6 nodes are sufficiently large to store all data including input, output, and intermediate data. How can we utilize storage effectively?
2. PageRank in Large Dataset In Lab 2, you worked with about 2,000,000 documents. Suppose you have 10,000 times more documents. What are the expected problems using PageRank algorithm and how can you solve those problems?
3. PageRankin Dynamic Dataset Web documents are created and removed as time goes. Computing PageRanktakes a large amount of time. How can we compute PageRankefficiently when graphs change all the time?
4. PaxosProject • What is “consensus”? • The process of agreeing on one result among a group of process (or a group of state machine) • What is Paxos? • Protocols for solving consensus in a network of unreliable processors • Paxos in wikipedia
Various PaxosProtocols • A family of Paxos protocols • Basic, Multi, Cheap, Fast, Generalized Paxos • A number of optimization skills • Reduce the number of messages between processors • Reduce the number of calculations in a process
Applications Using Paxos (I) • Google Chubby • Distributed lock service to keep replicas consistent • Used in Google Analytics and other products • http://labs.google.com/papers/chubby.html
Applications Using Paxos (II) • IBM SAN Volume Controller • General-purpose fault-tolerant virtual machine • Runs the configuration and control component of the storage virtualization services offered by the cluster • IBM SAN volume controller
Paxos Project • Project Objectives • Study the family of Paxos algorithms • Choose aspecific version of Paxos protocol and optimization skills • Design appropriate services or systems and implement it using the Paxosalgorithm • You have to simulate or make some failure conditions and prove the system’s fault-tolerance • Experiment over PlanetLab?
Schedule 10/30 Thu – Lab 2 review by TAs Class on Chapters 2 & 3 Individual team meetings Term project presentation #1 Class on Chapters 9 and 10 Individual team meetings (tentative) Class on Chapters 9 and 10 (tentative) Term project presentation #2 (tentative)