From best@cs.bu.edu Mon Mar 17 10:09:16 1997 Delivery-Date: Mon, 17 Mar 97 10:09:32 -0500 Return-Path: best@cs.bu.edu Received: from csb.bu.edu (best@CSB.BU.EDU [128.197.10.4]) by cs.bu.edu (8.8.5/8.8.5/(BU-S-01/27/97-fc1)) with ESMTP id KAA22291 for ; Mon, 17 Mar 1997 10:09:16 -0500 (EST) Received: by csb.bu.edu (8.8.5/Spike-2.1) id KAA09373; Mon, 17 Mar 1997 10:09:12 -0500 (EST) Date: Mon, 17 Mar 1997 10:09:12 -0500 (EST) Message-Id: <199703171509.KAA09373@csb.bu.edu> From: Azer Bestavros To: cs550@cs.bu.edu Subject: Next week I will be out of town on Monday March 24th. The lecture for that day will be offered on Friday March 28th at 4 in the same place. PLEASE TAKE NOTICE. Thanks, --Azer /\ Azer Bestavros - Assist Prof Comp Sci Dept ___ / \/\ Boston University, MA 02215 617-353-9726 /|__) _|_____ / \ \/\ Http://www.cs.bu.edu/~best best@bu.edu /_|__)es|avros / \ \ \_________________________________________/ |zer From best@cs.bu.edu Wed Mar 19 02:38:45 1997 Delivery-Date: Wed, 19 Mar 97 02:39:02 -0500 Return-Path: best@cs.bu.edu Received: from sphinx.bu.edu (best@SPHINX.BU.EDU [128.197.10.101]) by cs.bu.edu (8.8.5/8.8.5/(BU-S-01/27/97-fc1)) with ESMTP id CAA06769 for ; Wed, 19 Mar 1997 02:38:44 -0500 (EST) Received: by sphinx.bu.edu (8.8.5/Spike-2.1) id CAA15020; Wed, 19 Mar 1997 02:38:41 -0500 (EST) Date: Wed, 19 Mar 1997 02:38:41 -0500 (EST) Message-Id: <199703190738.CAA15020@sphinx.bu.edu> From: Azer Bestavros To: cs550@cs.bu.edu Subject: Assignments 2 and 3 are out! Roll up your sleeves! Assignments 2 and 3 have been posted on the Web. Please check them out. You may want to start working on assignment 2 ASAP. It deals with dynamic pipelining and is due in 9 days. Assignment 3 deals with advanced memory design issues and with Vector processing (the subject of our lecture tomorrow (oops, I mean today---it's almost 3am!) Assignment 3 is due in 21 days. --Azer /\ Azer Bestavros - Assist Prof Comp Sci Dept ___ / \/\ Boston University, MA 02215 617-353-9726 /|__) _|_____ / \ \/\ Http://www.cs.bu.edu/~best best@bu.edu /_|__)es|avros / \ \ \_________________________________________/ |zer From best Sat Apr 5 12:32:03 -0500 1997 From: Azer Bestavros To: cs550 Subject: Next few lectures and homework #3 Hi All: As I announced in the section meeting we had yesterday, there will be no class on WEDNESDAY 4/9 since I will be on my way to England for a conference, where I am giving a paper and chairing a session. Also, the next lecture on MONDAY 4/14 (a week from this coming Monday) will be given by Professor David Yates. I will forward the title and a short abstract about he will be talking about later today. Also, since there will be no class on Wednesday, the deadline for the homework will be extended till MONDAY 4/14. See you in class on Monday. --Azer From best Sat Apr 5 12:40:50 -0500 1997 From: Azer Bestavros To: cs550 Cc: djy Subject: Abstract for lecture on 4/14 Hi everybody: The following lecture by Professor David Yates will be offered one week from this coming Monday (on 4/14). It will be a very nice exposure on how experimentation is necessary to optimize memory performance. Obviously, David will be talking about work that is quite recent and that should give you a good exposure on a good example of cutting-edge research. For those of you who do not know of Prof. Yates, he joined our department last fall, teaches the data comm course and does research in networking (he is a member of the Oceans group that I briefly spoke about at some point in the course). --Azer Cache Behavior of Network Protocols David Yates http://www.cs.bu.edu/faculty/djy/ In this talk we present a performance study of memory reference behavior in network protocol processing, using an Internet-based protocol stack implemented in the x-kernel running in user space on a MIPS R4400-based Silicon Graphics machine. We use the protocols to drive a validated execution-driven architectural simulator of our machine. We characterize the behavior of network protocol processing, deriving statistics such as cache miss rates and percentage of time spent waiting for memory. We also determine how sensitive protocol processing is to the architectural environment, varying factors such as cache size and associativity, and predict performance on future machines. We show that network protocol cache behavior varies widely, with miss rates ranging from 0 to 28 percent, depending on the scenario. We find instruction cache behavior has the greatest effect on protocol latency under most cases, and that cold cache behavior is very different from warm cache behavior. We demonstrate the upper bounds on performance that can be expected by improving memory behavior, and the impact of features such as associativity and larger cache sizes. In particular, we find that TCP is more sensitive to cache behavior than UDP, gaining larger benefits from improved associativity and bigger caches. We predict that network protocols will scale well with CPU speeds in the future.