Chapter 17 Excerpted from the 1994 Research Summary
(Copyright 1994, UC Regents)

Contact person: Carol Block, ILP Coordinator
(ilp@hera.eecs.berkeley.edu)
510.643-6691, 643-6694 fax

Operating systems research is driven by technological developments and
user and application requirements. Among the most important
technological developments of the recent past are faster and more
reliable computer networks, the trend toward media digitalization and
integration, and the availability of more powerful and affordable
system and network components. These developments are making a number
of new applications feasible: for instance, those involving the
processing, storage, and transmission of digital continuous media such
as motion video and sound, those needing parallel and distributed
computations because of their performance or availability requirements,
and those demanding larger, faster, or safer storage than was
previously possible.

The projects described in this chapter address many of these
exciting technological and application challenges. The first
two abstracts are part of the NOW project (networks of
workstations), which is attempting to provide very high-speed,
low latency communication between workstations, so that they
can be used for massively parallel applications that have
traditionally run only on supercomputers. The next two
projects address issues in portable computing, including how
to reduce the power consumption of portable computer systems
and how to manage network communication as a machine moves
from place to place.

One of the most exciting systems projects at Berkeley is
Sequoia 2000, a collaborative effort between computer
scientists and earth scientists at many sites throughout
California. The goal of Sequoia 2000 is to bring the
expertise of the computer science research community to bear
on the problems of earth scientists, which include
transmission, storage, and analysis of massive amounts of
data. The next two abstracts in this chapter describe projects
to design an object storage system for Sequoia and to give
scientific application more control over the management of
their virtual memory.

The next group of abstracts concerns the Sprite and RAID
project, which have been developing new approaches to network
file systems and parallel disk arrays. These projects address
issues such as fast recovery after crashes, how to move data
from a disk array through a file server to a network at high
speed, how to extend the RAID ideas to stripe data across
multiple file servers, and how to deliver images from a disk
array over a network in real-time.

The last two abstracts in this chapter describe work to
analyze file caching techniques and a new approach to discrete
event simulation.