http:^^www.cs.utah.edu^~retrac^home.html
来自「This data set contains WWW-pages collect」· HTML 代码 · 共 343 行 · 第 1/2 页
HTML
343 行
Date: Thu, 21 Nov 1996 21:21:57 GMT
Server: Apache/1.1.1+
Content-type: text/html
Content-length: 13004
Last-modified: Sun, 07 Jan 1996 23:05:27 GMT
<html><head><title>John B. Carter</title></head><body><!WA0><img src="http://www.cs.utah.edu/~retrac/photo.gif"></a><hr><h1>John B. Carter</h1><b>Assistant Professor</b><br><b>Computer Science Department</b><br><b>University of Utah</b><br><br><hr>Professor Carter joined the Department of Computer Science in January 1993.His research interests include operating systems, parallel and distributedcomputing, and multiprocessor computer architecture. Of particularinterest are scalable shared memory architecture designs, both hardware andsoftware. Dr. Carter is co-leading two ARPA-sponsored research projects:the <!WA1><a href="http://www.cs.utah.edu/projects/avalanche">Avalanche scalable multiprocessorarchitecture</a> design effort and the <!WA2><a href="http://www.cs.utah.edu/projects/flexmach">Fastand Flexible Mach-Based Systems</a> effort (aka Mach 4). The developmentof an efficient and highly portable distributed shared memory system, <!WA3><ahref="http://www.cs.utah.edu/projects/flexmach/quarks.html">Quarks</a>, is part of this effort.<p>While a graduate student at Rice University, he designed, implemented, andevaluated <em>Munin</em>, a distributed shared memory system that allowsshared memory parallel programs to be executed efficiently on distributedmemory multiprocessors. He has also worked on <em> high speed bulk datacommunication protocols.</em><p><p>For the Fall Quarter of 1995, I will be teaching the <!WA4><ahref="http://www.cs.utah.edu/~cs506"> Introduction to Operating Systemscourse, CS506</a>. Meeting times are Tuesday-Thursday from 1:10pm-2:35pmin EMCB 102.<hr><dt><h2>Current Research Projects</h2><dl><!WA5><a href="http://www.cs.utah.edu/projects/avalanche"><!WA6><img alt = "o" src = "http://www.cs.utah.edu/icons/blueball.gif"></a><b>Avalanche Scalable Multiprocessor Design:</b>The goal of this project to develop an integrated cache, memory, andcommunication architecture that significantly reduces the latency of bothdistributed shared memory and message passage multiprocessor communication.The core processor of Avalanche will be the forthcoming PA-RISC 8000 CPU.We are designing a new <em>Context Sensitive Cache Controller Unit</em>that will support a flexible suite of cache coherence protocols for DSMapplications and provide context sensitive injection of incoming data intothe appropriate level of the memory hierarchy in order to minimize messagelatency.Primary research collaborators:<!WA7><a href="http://www.cs.utah.edu/~ald">Al Davis</a>,<!WA8><a href="http://www.cs.utah.edu/~kuramkot">Ravindra Kuramkote</a>, and<!WA9><a href="http://www.cs.utah.edu/~chenchi">Chen-Chi Kuo</a>.Other research collaborators:<!WA10><a href="http://www.cs.utah.edu/~swanson">Mark Swanson</a>,<!WA11><a href="http://www.cs.utah.edu/~stoller">Leigh Stoller</a>,<!WA12><a href="http://www.cs.utah.edu/~yih">Benny Yih</a>, and<!WA13><a href="http://www.sics.se/~ans/index.shtml">Ashley Saulsbury</a>.</dl><dl><!WA14><a href="http://www.cs.utah.edu/projects/flexmach"><!WA15><img alt = "o" src = "http://www.cs.utah.edu/icons/blueball.gif"></a><b>Fast and Flexible Mach-Based Systems:</b>The goal of this project is to develop an operating system that provides amuch higher degree of flexibility than traditional operating systems, andto use that added flexibility to circumvent the performance/functionalitytradeoffs that thwart traditional highly-decomposed, microkernel-basedoperating systems. Important components of this work are a modulemanagement service, lightweight and decomposed Mach kernel functionality,aggressive exploitation of interprocess sharing, and efficient distributedshared memory (see below). We will maintain backward compatibility wherepractical, and freely distribute an unencumbered version of the entiresystem.Primary research collaborators:<!WA16><a href="http://www.cs.utah.edu/~lepreau">Jay Lepreau</a>,<!WA17><a href="http://www.cs.utah.edu/~mike">Mike Hibler</a>,<!WA18><a href="http://www.cs.utah.edu/~law">Jeff Law</a>, and<!WA19><a href="http://www.cs.utah.edu/~baford">Bryan Ford</a>.</dl><dl><!WA20><a href="http://www.cs.utah.edu/projects/flexmach/quarks.html"><!WA21><img alt = "o" src = "http://www.cs.utah.edu/icons/blueball.gif"></a><b>Quarks Distributed Shared Memory System:</b>The goal of this project is to develop an efficient, portable, and freelyavailable distributed shared memory system to support the shared memoryprogramming style on distributed memory multiprocessors and networks ofworkstations.Ideally, Quarks will eventually do for distributed shared memory what PVMdid for message passing -- make it relatively easy to use, pervasive(ported to a wide array of systems), and reasonably efficient.For more on our motivation, take a look at my<!WA22><a href="http://www.cs.utah.edu/~retrac/hotos95.ps.Z">HOTOS '95 position paper</a>.An <!WA23><a href="http://www.cs.utah.edu/projects/flexmach/quarks.html"><b>Alpha release</b></a> of Quarks is currently available that runs on SunOS4.1/SPARC machines, but ports are in progress to 68K BSD boxes,HP-UX/PA-RISC, IRIX 5.2/MIPS, and the Mach operating system.Primary research collaborators:<!WA24><a href="http://www.cs.utah.edu/~khands">Dilip Khandekar</a> and<!WA25><a href="http://www.cs.utah.edu/~kamb">Linus Kamb</a>.<p>Here is a copy of the slides for my OSDI '94 <!WA26><a href="http://www.cs.utah.edu/~retrac/osditut.eps.Z">tutorial on distributed shared memory</a> minus some of the graphs,which I will incorporate soon.</dl><p><hr><h2>Significant Past Research Projects</h2><dl><!WA27><img alt = "o" src = "http://www.cs.utah.edu/icons/blueball.gif"><b>Munin Distributed Shared Memory System:</b>Munin was the first software distributed shared memory system to explorethe potential performance benefits of using a relaxed consistency model.Among Munin's novel features were a software implementation of the releaseconsistency model, the first <em>multiple writer</em> memory consistencyprotocol to address the problem of <em>false sharing</em>, support formultiple consistency protocols (including user-supplied protocols), and anupdate timeout mechanism to reduce the communication overhead of writeupdate protocols. For shared memory programs with moderate to high degreesof sharing, Munin achieved far greater performance (speedup) thanconventional distributed shared memory systems, usually within 10% of handcoded message passing performance. Many of the features and ideas firstdeveloped in Munin have appeared in subsequent DSM systems. Primaryresearch collaborators: <!WA28><ahref="http://www.cs.rice.edu/CS/faculty/willy.html">Willy Zwaenepoel</a>and <!WA29><a href="http://www.cs.rice.edu/CS/faculty/jkb.html">John Bennett</a>.Here is a copy of my <!WA30><a href="http://www.cs.utah.edu/~retrac/carter-thesis.ps.Z">dissertation</a> concerning Munin.</dl><dl><!WA31><img alt = "o" src = "http://www.cs.utah.edu/icons/blueball.gif"><b>Optimistic Bulk Data Transfer Protocol:</b>The key insight exploited in this effort was that when the first packet ina `blast' of bulk data is received by a node, it is highly likely that nopacket outside the `blast' will arrive before the last bulk data packet isreceived. As such, upon receipt of the first packet in a blast of bulkdata, the network device layer should set things up to optimize this case.This optimization, which was independently observed by Van Jacobson andincorporated into TCP/IP, resulted in 9.2 Mbps bulk data transfer ratesbetween two SUN-3/50's on a 10 Mbps Ethernet. At the time, the best TCP/IPimplementations achieved under 5 Mbps bulk data transfer rates.Primary research collaborators:<!WA32><a href="http://www.cs.rice.edu/CS/faculty/willy.html">Willy Zwaenepoel</a>.</dl>⌨️ 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?