CS 525: Advanced Distributed Systems, Spring 2009, UIUCCS 525: Advanced Distributed Systems(previously CS 598IG)Course OverviewOver the past few decades, the functions of “traditional” operating systems have been scatteredout to the edges of distributed systems. Peer-to-peer systems (think: Kazaa), sensor network s, theGrid, PlanetLab, the Inter n et and the Web are examples. New classes of these systems includedatacenters and comput ing clouds (e.g., EC2, Appengine, Cirrus, Goog le-I BM cloud). This coursefocuses on two case study areas, along with a new emerging area: peer-to-peer systems, sensornetworks, and data-intensive cloud computing. We will study efficient protocol design and eval-uation, as well as learn h igh -level system issues. Research in these areas also tend s to be scatteredacross disjoint sets of researchers and conferences. A second emphasis of t h e course is on bridg-ing the gap between thes e focus areas and the existing base of theoretical d istributed computing.This has the potential for applying algorithms or principles from one area to another; the courseis an attempt to highlight these possibilities and motivate valuable projects.PrerequisitesBasic Computer Science and basic compu ter programming skills are essential. Knowledge of Op-erating Sy stems (CS 241 or CS 423), or Networks (CS 438), or Distributed Systems (CS 425), or anequivalent course, o r instructor consent, is required.Course Website http://www.cs.uiuc.edu/class/cs525.TimingsClass: Tuesday and Thursday, 9:30 AM - 10:45 AM, 1302 Siebel Center.Office Hours (tentative, class d ays only): Tuesday and Thursday 10.45 AM - 12 PM, 3112 SC.Course StaffDr. Indranil Gupta3112 SC, [email protected], 265-5517Course ContentThe first few weeks o f lectures introduce ground basics in peer to peer systems, theory, sensor net-works, and cloud computing systems. Subsequently, over 70 research papers in various areas ofdistributed systems are presented, discussed, and debated by the students. The selection includesclassical and contemporary papers from conferences including, but not limited to, PODC, Mid-dleware, SOSP, OSDI, Usenix, Infocom, SIGCOMM, SASO, etc., as well as ACM and IEEE journalpapers. Rather than running through all the papers in a few of such proceedings and journals, wewill pick and choose publications appropriate to the stated goals of this course.An e ssential component of the course is a project involving at least one non-trivial idea andhands-on implementation. If this project leads to a conference-quality paper submission/acceptance,your course grade could benefit from it. You can collaborate in groups, and I will work with indi-vidual groups in defining the project and during its progress. At the end of the course, a top few1“best” project papers will be selected, and given special attention for submission to conferences.For the record, 9 out of the 12 papers from the Fall 2003 course have been accepted in conferences(e.g., WCW, ICDCS, MASS and PODC), and many of the F all 2004, Spring 2006, Spring 2007, andSpring 2008 papers are either under conference review or have been accepted into highly comp et-itive conferences (e.g., ICDCS, Infocom, Middleware, MMCN, SASO, QShine, CollaborateCom,GRID, e tc.), as well as appear in several top journals (e.g., ACM TAAS, ACM TOSN, IEEE TNSM,JSS, Distr ibuted Compu ting, etc.).About the ClassThe initial few we eks of class will consist of lectu r ing , with the intent of building up commonknowledge and grounding for the latter half of the course. We will then transition to student-ledpresentations of papers. Once student-led presentations start, students who are not presenting areexpected to write short reviews (1-2 pages total) for any two of the “Main Papers” in that s ession.Active class participation is required, even in t h e initial part of the course!Class Evaluation: Project, review papers, p resentation, and class participation. Tentative splitsare 50%, 20%, 25%, 5% respectively.Abridged list of Topics (see course website for more comprehensive list): Distributed comput-ing theory, probabilistic algorithms, peer to peer syst ems, cloud computing, sensor networks, theGrid, overlays, routing, handling stress, distributed management, data-intensive programming,membership, classical algorithms, design methodologies, sources of unreliability and trace stud-ies, industrial systems, caching, publish-subscribe, structure of networks, selfish
View Full Document
Unlocking...