http:^^www-cse.ucsd.edu^users^marzullo^
来自「This data set contains WWW-pages collect」· EDU^USERS^MARZULLO^ 代码 · 共 114 行
EDU^USERS^MARZULLO^
114 行
Date: Thu, 21 Nov 1996 20:09:20 GMTServer: NCSA/1.4.2Content-type: text/htmlLast-modified: Tue, 22 Oct 1996 16:58:04 GMTContent-length: 4437<HTML><br><br><TITLE>Keith Marzullo's Home Page</TITLE><H1><!WA0><a href="http://www-cse.ucsd.edu/users/marzullo/marzullocanoe.html"><!WA1><IMG SRC="http://www-cse.ucsd.edu/users/marzullo/face4.gif" width=200 height=200></A></H1> <H3>Keith Marzullo<br>Department of Computer Science and Engineering<br>University of California, San Diego<br>9500 Gilman Dr. 0114<br>La Jolla, CA 92093-0114<br></H3><P><B>Office:</B> <!WA2><A HREF="http://www-cse.ucsd.edu/users/marzullo//cse/APM.html"> APM</A> 4220<br><B>Phone: </B> (619) 534-3729<br><B>Fax: </B> (619) 534-7029<br><B>Lab: </B> (619) 534-9669<br><B>Email: </B> <!WA3><A HREF="mailto:marzullo@cs.ucsd.edu">marzullo@cs.ucsd.edu</A><br><B>Secretary: </B> <!WA4><A HREF="http://www-cse.ucsd.edu/users/marzullo//users/mfoley/">Michele Foley</A> (619) 534-5228<br><P><!WA5><IMG SRC="http://www-cse.ucsd.edu/users/marzullo/line4.gif"><P>I'm an associate professor and do <!WA6><A HREF="http://www-cse.ucsd.edu/users/marzullo/bibliography.html">research</A> in fault-tolerant distributed systems. My most recentinterests are in the foundations of group-based programming and itsapplication to distributed computing architectures different than a collection of asynchronous workstations running on local area networksof workstations. I'm affiliated with the following research projects:<UL><LI> <!WA7><A HREF="http://www.nile.utexas.edu/NC/"> The Nile Project</A>that has to do with distributed computing for high-energyphysicists. Our goal is to provide a deep hierarchical storage systemand a distributed runtime environment for very long-running jobs. Mymain focus on this project is the distributed application managementissues: I'd like to be able to have the system adapt to changes inresources, communication platforms, internet connections and internetcharges without having to rewrite the system or the applicationsrunning on the system.<br><br><LI> <!WA8><A HREF="http://www-cse.ucsd.edu/users/marzullo/team.html"> The Team Project</A> that is buildingdistributed computing platforms for both hard and soft-real-timesystems. My main focus has been on providing the runtime environmentto make fault-tolerant hard real-time computing available toinexpensive platforms.</UL>Some things that I'm working on with students and ex-students:<UL> <LI> The message logging approach to making services tolerate crashand communication failures, with <!WA9><AHREF="http://www.cs.utexas.edu/users/lorenzo/">Lorenzo Alvisi</A>. We've come up with a simple and expressive way tocharacterize the approach and have derived practical and optimalprotocols for both message passing and distributed shared memoryarchitectures. A paper describing the approach can be found <!WA10><A HREF="http://www-cse.ucsd.edu/users/marzullo/dcsfinal.ps"> here</A>and a paper giving a set of protocols (with a comparison to Manetho)can be found <!WA11><A HREF="http://www-cse.ucsd.edu/users/marzullo/ftcs.ps">here</A>.<br><br><LI> Group programming in a hard real-time environment, with <!WA12><AHREF="http://www-cse.ucsd.edu/users/marzullo//users/mclegg/"> Matthew Clegg</A>.We've combined some of the schedulability work done for uniprocessorswith the process group abstraction for distributed systems to come upwith a system that can tolerate timing, crash and communicationfailures while still meeting hard real-time deadlines.<br><br>One of the problems in such systems, of course, is clocksynchronization. A paper that describes how we do this can be found<!WA13><A HREF="http://www-cse.ucsd.edu/users/marzullo/clocks.ps"> here</A><br><br>.<LI> Failure models, with <!WA14><AHREF="http://www.cs.cornell.edu/Info/People/sabel/sabel.html">Laura Sabel</A>. We've taken apart some of the group membershipprotocols to better understand what are the lower bounds are forseveral of the properties that are commonly provided. One way to thinkof this work is that we're looking at failure detectors that are <em>stronger</em> than a Perfect Failure Detector and at<em>approximations</em> of such failure detectors that areimplementable.<br><br>If you're interested in failure detectors in general, you might takea look at <!WA15><AHREF="http://www.cs.cornell.edu/Info/People/chandra/FailureDetectionPapers.html">a WWW page</A> that compiles papers on the topic.<br><br><LI> Comparison of fault-tolerance techniques,with <!WA16><AHREF="http://www-cse.ucsd.edu/users/marzullo//users/jsussman/"> Jeremy Sussman</A>.The goal here is to understand what are the intrinsic benefits of agiven fault-tolerance paradigm, such as the primary-backup approach,coordinatory-cohort, or active replication.</A></UL><!WA17><IMG SRC="http://www-cse.ucsd.edu/users/marzullo/line2.gif"><!WA18><A HREF="http://www-cse.ucsd.edu/users/marzullo/dancing.html"><!WA19><IMG SRC="http://www-cse.ucsd.edu/users/marzullo/dance.gif"></A> Here's a link for Aleta and Michael. <br> <br><!WA20><A HREF="http://www-cse.ucsd.edu/users/marzullo//CSE_Home_Page.html"> Back to the departmentalhome page</A>.<PRE> last revised 19 January 1996 </PRE> <!WA21><IMG SRC="http://www-cse.ucsd.edu/users/marzullo/bike_cartoon.gif">⌨️ 快捷键说明
复制代码Ctrl + C
搜索代码Ctrl + F
全屏模式F11
增大字号Ctrl + =
减小字号Ctrl + -
显示快捷键?