[13] in Open-Software-Foundation-News
OSF Flash Bulletin - OSF RI Mission and Plans
daemon@ATHENA.MIT.EDU (OSF Corporate Communications)
Wed Mar 9 13:36:42 1994
Resent-From: Bill Cattey <wdc@MIT.EDU>
Resent-To: osf-news-mtg@menelaus.LOCAL
To: newsnug@osf.org
Reply-To: newsnug@osf.org
Date: Tue, 08 Mar 1994 20:53:54 -0500
From: OSF Corporate Communications <corpcom@osf.org>
To: OSF Members
From: corpcom@osf.org (OSF Corporate Communications)
Open Software Foundation
****************************************************
OSF ELECTRONIC FLASH
****************************************************
An electronic mail news flash for OSF Members from
the Open Software Foundation
March 8, 1994
<><><><><><><><><><><><><><><><><><><><><><><><><><><><>
OSF Research Institute Mission and 1994 Project Plans
<><><><><><><><><><><><><><><><><><><><><><><><><><><><>
PREAMBLE
Reproduced below (in ASCII format suitable for News
Nuggets) are two OSF Research Institute documents, "The
OSF Research Institute Mission Statement" and "RI 1993
Project Summary and 1994 Plans". Both these documents
are available in hardcopy with better formatting than is
possible here.
Together these documents give a good overview of the RI
as we enter 1994.
THE OSF RESEARCH INSTITUTE MISSION STATEMENT
By Dr. Ira Goldstein
Vice President,
Research and Advanced Development
February 25, 1993
Updated October 30, 1993
1. Introduction
The OSF Research Institute (RI) is chartered to
investigate fundamental software technology needed to
improve the scalability, portability and interoperability
of computer systems. We conduct these investigations
collaboratively with academia and industry. Our
methodology is to develop complete prototypes that can be
subjected to rigorous measurement as well as provide a
platform for others to build on our work. We are
successful if our research delivers innovation and
economy: innovation by providing solutions to difficult
system technology requirements and economy by providing a
vehicle for shared investment in this research.
Section 2 presents our technology focus and section 3
describes our methodology.
2. Technology Focus
Open systems are characterized by scalability of a common
system technology from single processor hardware
platforms to massively parallel systems, portability of
applications to any of these systems, and
interoperability among these systems. Our technology
investigations are directed at enhancing all three of
these properties.
Scalability. There is a revolution underway in hardware
parallelism that has the potential to deliver
breakthroughs in performance and price/performance.
However, fundamental improvements in operating system
technology are needed. Hence, we are developing a new
system technology architecture that can make effective
use of parallelism. Our goals include scalable
performance as well as attention to critical system
properties such as real-time response,fault tolerance,
high availability, and high bandwidth I/O interactions.
Our results are available as experimental systems, which
can be used in toto or from which technology can be
extracted.
Portability. End users must be able to choose their
hardware without obsoleting their software if ongoing
innovation in hardware design is to be practical and end
users are to benefit fully from a competitive
marketplace. Present software distribution formats are
tied to particular hardware architecture/operating system
combinations. To break through this barrier of software
lock-in, we are participating in an international program
to develop ANDF. ANDF is an architecture-neutral software
distribution format that can span multiple languages and
multiple architectures. This technology is fundamental to
application portability and complements the development
of a open systems standards. It also simplifies the
provision of compilers for new programming languages. The
RI's special focus is on advanced validation techniques
and on extensions to parallel languages. Other projects
sponsored by Esprit are developing this technology for
conventional languages (C, C++, Cobol, Fortran, Lisp) and
new hardware architectures. The Grenoble Research
Institute is acting as a bridge between the European and
U.S. work in this area.
Interoperability. The RI is investigating the development
of system technology that can integrate small and large
systems on a local area network into a single coherent
computing environment. Such an environment provides
efficient utilization of idle computer cycles, simplified
system administration, increased availability and the
potential for high performance parallel computing at a
reasonable price. We are adapting our scalable, multi-
computer technology to provide this environment within
the less reliable context of a local area network. This
technology is usable on today's networks but becomes
especially cost-effective as networks are upgraded to
high speed fibre optics.
Usability. The enormous information and computational
capacity of the emerging worldwide information
infrastructure will not be of value unless we can develop
intuitive user interfaces that work well in distributed
environments. The Mosaic technology from University of
Illinois is an example of this new generation of
hypermedia browsing interfaces to information distributed
on the Internet. The RI is initiating a new program whose
goal is to increase the usability of this technology by
designing intelligent assistants that can guide the user
in his investigation of network information. Proposals
have been submitted to various government programs to
augment the RI's available funding for research in this
area.
Our present focus is on the impact of parallelism on
fundamental system capabilities and on the opportunities
that the emerging distributed information infra-structure
presents to users. We will investigate other open systems
technologies as resources permit. Overall our goal is to
prototype advanced technology that accelerates and make
more economic the delivery of new capabilities to the
community.
3. Methodology
Our methodology emphasizes collaboration. This is always
an effective and economic working style, but it is
especially appropriate for research concerned with open
systems. We work with leading universities and cutting
edge companies. For example, we selected Carnegie
Mellon's Mach technology as a starting point in 1989 and
have worked closely with them in the ensuing years. This
collaboration has proved successful, with the results
being transferred to OSF Engineering and directly to
over 40 sites. We are especially pleased to collaborate
with early adopters. Recently, this has included
companies developing a new generation of massively
parallel systems such as Convex and Intel. Our technology
is also being explored by various vendors as a foundation
for multi-personality, low-end operating environments,
for high performance, real-time systems, and for
clustered computing.
The RI develops complete system prototypes. RI prototypes
synthesize code contributed by universities and leading
edge companies as well as our direct efforts. They are
sufficiently complete so as to reveal the effects of
various architectural elements as systems are tested
under conditions of significant load (where load may
represent the number of processors, the size and
complexity of applications, the number of simultaneous
tasks, or other stress imposed by the requirement that
open systems technology be scalable). Several companies
have used RI prototypes to accelerate their product
schedules.
The RI has an international perspective. Research
benefits from cross-fertilization. The RI collaborates
with partners worldwide and makes its technology
available on a worldwide basis. About one third of the
RI's 60+ staff members are located in Grenoble, France
and the remainder are based in Cambridge, Massachusetts.
The RI regularly presents its work at symposia organized
for OSF members in Japan, Europe and the U.S. each year.
It also acts as a communications channel between these
various communities. The RI just hosted a Symposium in
Cambridge which was attended by over 90 engineers and
managers from the OSF sponsor companies.
The RI is a highly leveraged investment. The RI has
funding from major computer companies, contracts with
government agencies (DARPA and Esprit), equipment grants
from vendors, and direct support from OSF. These
investments are synergistic. Each company's investment
isleveraged by at least a factor of 10.
RI methodology emphasizes relevance. Guidance from
sponsor companies, OSF members, government funding
agencies and licensees insures that RI activities are
focussed on critical problems and that transfer of
completed research is facilitated. The RI is not an ivory
tower. Typically RI staff visit sponsor companies at
least once each quarter and interact with other member
companies at their request.
The RI produces several kinds of deliverables. We produce
analyses, measurements, architectural frameworks and
system prototypes. These results are delivered to the OSF
membership and the OSF Engineering organization. Our
prototype code may be used directly by OSF members,
integrated into offerings from OSF Engineering, or serve
as a starting point for OSF technology procurements.
Alternatively, our analyses are also valuable when they
provide insight into software techniques that are not
yet ready to become products.
RI results are open. Analyses are reported to the
technical community and RI prototypes are available to
OSF members as unsupported code. The RI places no
limitations on the use of its prototype code in products
or as the basis for further research beyond those imposed
by whatever licenses were required for the RI to develop
this software. The RI hosts several workshops each year
and presents its work regularly at conferences.
4. Conclusion
As the rate of change in the computer industry
accelerates, there will be an ongoing need to deliver
innovative products rapidly and economically. It is the
RI's mission to serve this need through a cost-effective
program of shared investment and open results.
-------------------------------------------------------
RI 1993 PROJECT SUMMARY AND 1994 PLANS
By Dr. Ira Goldstein
December 4, 1993
I. OPERATING SYSTEM FUNDAMENTALS AND MULTIPLE
PERSONALITY MACH
A. Program, Purpose, Collaborators
Achieve reduced cost and increased functionality by
providing a common foundation for open and proprietary
operating systems.
DEC, HP, IBM, Hitachi, Convex, Intel, Utah, ARPA
B. Projects
OSF/1 MK: Develop a generic server-based operating system
architecture that can support multiple personalities.
Provide techniques such that performance is within 95% of
an integrated architecture.
Develop a Mach specification that unifies input from
industry and academia.
Explore xKernel from U. of Arizona as a framework for
defining network protocols.
Integrate the latest Mach improvements from CMU.
C. 1993 Accomplishments
No-emulator prototype (OSF/1 MK r5) completed to improve
reliability & trust and transferred to Eng. as basis for
OSF/1 r1.3.
Unified specification transferred to Eng. as basis for
r1.3 product. Assistance provided to Eng. so that r1.3
will be conformant.
Untyped IPC & multiprocessor extensions completed and
transferred to IBM, Convex, Intel and others.
Port to DEC AXP.
Microkernel benchmark suite completed and transferred to
IBM, Convex and others.
Preliminary xKernel implementation completed.
Latest improvements and bug repairs from CMU integrated
into MK5 and transferred to IBM and others.
D. 1994 Goals & Plans
Design and prototype performance enhancement techniques
such as server co-location. Achieve within 5% of the
integrated kernel. Release the improved system as OSF/1
MK r6 in 2Q94 on Intel and HP-PA.
Demonstrate generality by porting the HP-UX compatibility
package to MK6.
Analyze IBM MK, propose improvements and assist OSF
sponsors in their evaluation of this technology.
Measure performance and ease of maintenance of network
protocols defined via the x-kernel technology.
II. REAL-TIME MICROKERNEL ENHANCEMENTS
A. Program, Purpose Collabors
Providing real-time characteristics for distributed
systems is a new and challenging system technology
requirement for the 90's.
IBM, DEC, Center for High Performance Computing, ARPA,
HiPer-D sites, Rome Lab, Honeywell, Hughes.
B. Projects
R/T Enhancements: Improve clocks, timers, inter-process
communication.
Make kernel preemptive.
Design new algorithms for local and distributed IPC.
Prototype r/t synchronizers.
C. 1993 Accomplishments
Releases of R/T technology transferred to IBM, DEC, ARPA
HiPer-D sites and Eng.
Collaboration initiated with DEC. Three part-time
sabbaticals began Dec. 1.
Contract awarded by Air Force to integrate advanced real-
time from the Alpha operating system into Mach.
D. 1994 Goals & Plans
Design and prototype kernel preemption, improved latency,
local and remote R/T IPC, synchronizers and threads,
characterization tools, and scheduling framework for
separating mechanisms from policies.
Release MK 7 in 4Q94 & interim release 2Q94.
Integrate 1st version of distributed threads into Mach.
III. MASSIVELY PARALLEL EXTENSIONS TO MACH AND TO OSF/1
A. Program, Purpose Collagorators
Massively parallel processing is becoming an economic
means to supply high end computing needs. New O/S
technology is needed to realize the potential of these
systems.
CMU, Intel, DEC, IBM, Locus, Hitachi, ARPA, Univ. of
Arizona, Princeton, Honeywell.
B. Projects
OSF/1 AD: Develop a distributed operating system that
provides a single system image efficiently over 1000's of
nodes. Modify Mach's mechanisms to extend seamlessly to
an MPP system. Extend the VM system to simulate shared
memory efficiently. Allow all OSF/1 services (file
system, process management, networking) to be distributed
transparently. Scale to 100's and then 1000's of
processors.
C. 1993 Accomplishments
OSF/1 AD r1.0 completed and provided to interested
members. Assistance provided to early adopters such as
Intel and Convex. Intel shipped their 1st release in
April, 93.
Preserve symmetric multiprocessing capability for nodes
in the MPP system.
D. 1994 Goals & Plans
OSF/1 AD r2.0 planned for Dec. 94 leveraging the MK6 code
base. Provide an improved, redesigned distributed IPC
(NORMA IPC), and extensible distributed file system and
process management framework.
Provide a mid-year interim release of OSF/1 AD that
includes a partial implementation of the new NORMA IPC
facility. This release is needed by early adopters.
IV. CLUSTERS
A. Programs, Purpose Collaborators
End users will benefit from the ability for their many
workstations to work together seamlessly as a high
performance cluster. This is especially valuable as
higher speed networks are introduced.
HP; Convex; Trinity College, Dublin; Univ. Catholique de
Louvain, Belgium; Kaiserslautern Univ., Germany
B. Projects
The OSF/1 AD system was developed on PC clusters, but
assumed a reliable connection. Key mechanisms like IPC
must be made fault tolerant. Also it must be enhanced to
use DCE where appropriate.
C. 1993 Accomplishments
A fault-tolerant architecture for OSF/1 AD has been
designed, which includes Mach IPC extensions which will
allow shutdown and reboot in case of node failure.
Enhancements have been made to the network protocols used
by Mach IPC, which are essential for efficient operation
of OSF/1 AD over clusters.
The general problem of cluster configuration (both at
boot time and dynamically in case of node failure) has
been studied.
D. 1994 Goals & Plans
Much of the AD r2.0 work can be leveraged in this space.
Implement node failure detection, shutdown and reboot for
OSF/1 AD on a cluster.
Implement the required Mach IPC enhancements required for
cluster configuration. Develop boot configuration and
reconfigurationmechanisms for OSF/1 AD on clusters.
V. HIGH TRUST
A. Programs, Purpose Collaborators
Greater trust in computer systems will be an important
theme in the 90's as we rely more and more on distributed
machines outside our organizational control for our
information.
Trusted Information Systems, NSA, Convex, ARPA, HP
B. Projects
Rearchitect Mach to meet high trust requirements. The
emphasis is on producing a carefully designed and layered
implementation of the Mach kernel in C++ which will
achieve the high level of assurance of correct operation
demanded at the B3 level. This implementation should
equally appeal to the quality conscious computer
industry.
C. 1993 Accomplishments
Completed first step: Produced a functional definition of
OSF Mach. Published for member review and adopted by OSF
Eng.
Published a multi-layer structure for the rearchitected
kernel.
Produced a C++ reimplementation of Mach IPC.
D. 1994 Goals & Plans
Publish a complete High Level Design for the high trust
kernel in May 94 for submission to NSA.
Develop reimplementation of OSF Mach in C++ for release
in 1995.
Explore B1 trust compliance of microkernel system using
existing OSF/1 technology.
VI. ANDF
A. Program, Purpose Collaborators
ANDF makes portability a reality for the end user.
DRA, Esprit partners, USL, HP, Microsoft, ARPA
B. Projects
Cooperative research with Esprit partners including DRA
and 6 other European companies.
Collaborative testing of major ISV applications with
support from HP and USL.
Development of a research kit.
C. 1993 Accomplishments
Design documents for ANDF validation suite released to
Esprit partners.
Oracle, Mosaic, Postgres ported and tested.
Techniques to use an existing compiler (GCC) as ANDF
installer developed.
Results published in ANDF Collected Papers, volumes 1, 2
and 3.
D. 1994 Goals & Plans
Complete validation suite & begin design of ANDF
interpreter.
Benchmark new releases from DRA.
Use ANDF on a regular basis for Mosaic research.
Extend ANDF to parallel systems. (ARPA)
Release ANDF kit with DRA producer, GANDF installers and
DRA installers.
VII. DISTRIBUTED INFORMATION ENVIRONMENT
A. Program, Purpose Collaborators
The emerging worldwide information infrastructure will
provide significant productivity advantages to knowledge
workers of the mid to late 90's providing that the
technology is easy to use.
Research partners have been identified in the various
proposals submitted to the government. In addition,
visits are underway to the sponsor companies to identify
interest therein.
This program is contingent on one of the submitted
proposals being funded.
B. Projects
Explore ease-of-use and productivity value of the Mosaic
Internet browsing technology.
Explore agent paradigm for intelligent assistance for
Internet Browsing.
Explore integration of DCE with Mosaic.
Employ healthcare as an experimental domain.
C. 1993 Accomplishments
Mosaic compiled with ANDF and used on an exploratory
basis.
RI World Wide Web server created with Microkernel and
ANDF documentation.
Four Proposals submitted:
1. Building a Scalable America with National
Consortium of High Performance Computing sites. (ARPA
TRP)
2. Exploring a medical information infra-structure
with Univ. of MA Medical Center. (ARPA TRP)
3. Intelligent Assistance for Internet Browsing with
advisory board of leading AI researchers. (ARPC SISTO
office).
4. White paper on integration of DCE with Mosaic.
(Rome Lab)
D. 1994 Goals & Plans
Collaborate with interested parties in sponsor companies
on exploration of the benefits and limitations of the
Mosaic technology for browsing hypermedia information on
the Internet.
Carry out research described in proposals if funded.
Use proposal dollars to create a leadership team for
research and adv. dev. in distributed computing and
information access.
The Open Software Foundation is a not-for-profit research and
development organization headquartered in Cambridge,
Massachusetts with more than 400 members worldwide.
# # #
OSF/1, Motif, the OSF logo, OSF, and Open Software Foundation
are trademarks of the Open Software Foundation, Inc.
Copyright 1994, Open Software Foundation, Inc. All
rights reserved.
*****************************************************
Donna Esterling, Editor Promotional Programs Mgr
Open Software Foundation email: newsnug@osf.org
11 Cambridge Center phone: 617-621-8700
Cambridge, MA 02142 USA fax: 617-621-0631
c/o OSF Electronic Bulletin
Comments to the editor can be sent to:
newsnug@osf.org
E-mail additions and changes can be sent to:
bulletin-request@osf.org
*****************************************************
