For each class meeting one hard-core technical paper is assigned. In
addition, some background reading is suggested. You are expected to read the
paper assigned in preparation of the class meeting. The student who presents
the paper is also required to do some of the background reading. Other students
are strongly encouraged to do at least a cursory read of the background
material.
Electronic versions are linked where available; see the course secretary for
paper copies. We will try to handout hardcopies of the assigned reading a week
in advance in class.
For background read Computer Architecture: a quantative approach (2nd ed), by Hennesey and Patterson, Morgan Kaufmann publishers.
For additional motivation behind this paper checkout the Network of Workstation project. In
particular, The case for
NOW.
This paper uses the SimOS environment for simulating the architectural
trends. Embra is the key
component of the system that makes the SimOS environment fast enough so that
real systems can be simulated. Since this paper is all about modern
uniprocessors and multiprocessors, it might be helpful to check out Computer
Architecture: a quantative approach (2nd ed), by Hennesey and Patterson,
Morgan Kaufmann publishers.
Another interesting performance study is The
Measured Performance of Personal Computer Operating Systems
J.B. Chen, Y. Endo, K. Chan, D. Mazieres, A. Dias, M. Seltzer, M.D.
Smith (Harvard). It uses the performance counters of the Pentium to study
applications running on NetBSD, Windows95, Windows/NT.
There is quite a bit of current discussion on what an appropriate
organization for an operating system is. This paper discusses one
approach. Next week's paper discusses yet another approach. The On Micro-Kernel
Construction paper, which also appeared in SOSP-15, argues the
microkernel organization is just fine; one has to implement them right.
This paper discusses the design and implementation of a switch for an
ATM-style network. Another good paper that describes a complete ATM-like
network is Rodeheffer, Edwin H. Satterthwaite, and Charles P. Thacker. Autonet:
A high-speed, self-configuring local area network using point-to-point
Links. IEEE Journal on Selected Areas in Communications 9, 8, (October 1991)
pages 1318-1335. If you haven't read the Ethernet paper, you should read it:
Robert M. Metcalfe and David R. Boggs. Ethernet: Distributed packet switching
for local computer networks. Communications of the ACM 19, 7 (July, 1976) pages
395-404.
Some papers about ATM are: ATM Concepts, Architectures, and Protocols Ronald
J. Vetter ATM Network: Goals and Challenges and B. G. Kim and P. Wang
Communications of the ACM, Vol. 38, No. 2, February 1995.
If you are unfamiliar with the Internet protocols, you should read The Design
Philoshophy of the DARPA Internet Procotols David D. Clark, Proceedings of the
1988 SIGCOMM Symposium, pp 106-114, Stanford, CA, August 1988. Another good
paper is An Analysis of TCP Processing Overhead David D. Clark, Van Jacobson,
John Romkey, and Howard Salwen, IEEE Communications Magazine, Vol. 27, No. 6,
1989.
This paper describes a radically-different architecture for structuring
networks: protocols are replaced by programs. As background reading you should
read Architectural Considerations for a New Generation of Protocols, D.D. Clark
and D.L. Tennenhouse, Proc. SIGCOMM 90 Symposium on Communications Architectures
and Protocols, Philadelphia, PA, Sept. 1990, 200-208.
Another timely topic: mobile code. A different approaches are explored to run code safely in another domain. The destination domain can either be the kernel (as in SPIN, Exokernel, or Vino) or be another processor (as in Java and Omniware). The different techniques used to ensure safety are sandboxing (ie., software fault isolation), type-safe languages (SPIN and Java), and self-certifying code.
The papers in the previous class meeting demonstrate the need for security
and discusses techniques. This paper describes a complete design and
implementation of security in a distributed system. It uses authentication
logic to reason about the security properties of the system. Check out
"Authentication in distributed systems: theory and practice" by Lampson et al.,
ACM Trans. on Computer Systems, vol 10, no 4, Nov. 1992, pp 265-310 for the
details on authentication logic.
This paper descibes what it takes to build an operating system capable of
receiving an A1 rating. The various ratings are described in the Orange Book.
Petal is an ideal storage system with unlimited capacity, always available,
and easy to manage. Petal is an alternative to RAID: Randy H. Katz, Garth
A. Gibson, and David A. Patterson. Disk system architectures for high
performance computing. Proceedings of the IEEE 77, 12 (December, 1989) pages
1842-1857.
The literature on file system is huge. The XFS file system is a state-of-the
art file system supporting large files, large number of files, large
directories, and high performance I/O. The background reading for this paper is
Fast File System (FFS), McKusick, M. Joy, W. Leffler, S. Fabry, ACM Trans. on
Computer Systems, Vol. 2, No. 3, Aug. 1984, pp. 181-197.
Serverless is a total misnomer, but the paper describes a radical approach to
building distributed file systems. The background reading for this paper is
RAID (see above), LFS (Mendel Rosenblum and John K. Ousterhout. The design and
implementation of a log-structured file system. Proceedings of the Thirteenth
ACM Symposium on Operating Systems Principles, Pacific Grove, California October
13-16, 1991, pages 1-15), and Zebra (The Zebra Striped Network File System,
J.K. Hartman and J.K. Ousterhout. Proceedings of the Fourteenth ACM Symposium on
Operating System Principles, Asheville, California, pp. 29-43). The prototype
serveless file system is called xFS, not to be confused with XFS (see
above).
This paper is the first of two paper on global memory management from the
University of Washington. The follow-on on this paper appeared last month at ASPLOS and discusses an
optimization on the first paper. The optimization involves sending subpages
instead of sending complete pages.
This paper presents a novel approach in the context of object-oriented
databases on the many issues that showed in 6.853. Check out Thor for more information.
The topic for November 13 is searching in the Internet. A hot and fast
moving area of research. To cover some of the different issue two
light-weight papers are assigned. The first paper is one of the Harvest
papers. The other paper is one of the first papers on Web crawlers. Of
course, you are encourages to check out other search tools, such as Alta
Vista, Yahoo, etc.
This is the first of two class meetings on mobile computing. We will
first take an application perspective on mobile computing by studying Rover. Rover was inspired and
builds on pioneering work done at CMU on Coda, a
file system for disconnected computing. Mobile computing is a hot and
growing research area; recently ACM started a new conference Mobicom on this
topic.
In this class meeting we will take a networking perspective on mobile
computing. In particular, we will focus on Mobile IP, a new protocol to
make IP work when hosts are moving. Mobile IP is in the process of being
standarized by the Mobile IP
working group.
Early on the term we read about multiprocessors that provided
distributed shared memory in hardware. This paper reports on the latest
advance in implementing distributed shared memory on software. There are
zillion of other papers on software distributed shared memory. It is
probably the most popular PhD. thesis topic in systems of the last decade.
The Shasta work references many of these other projects; feel free to
check them all out.
A recent paper from DEC SRC on distributed network objects. Most of the
issue discussed in this paper appear are at the hart of other systems, such
as distributed OLE-2 (or Active-X), Java's Remote Method Invocation
package, CORBA, etc. This paper is different from all the hype in that it
gives a crisp and clear description of what the problems are, how these
problems were solved, and what the performance issues are.
Dynamic code generation is a recent popular topic. This one of 3 papers
that appeared at the last PLDI on this topic.
A cool paper to make video conferencing over the Internet better.