From: d...@spsd.SPSD.DG.COM (Dennis D. Sherod)
Subject: DG/UX -- Seybold's Review ( long )
Date: 12 Sep 88 02:02:26 GMT
Reply-To: d...@sdsa03.UUCP ()
Organization: Data General SPSD Santa Ana, CA
The following is reprinted with permission.
EXCERPTS FROM: "DATA GENERAL'S UNIX STRATEGY: AN EVALUATION"
By Judith S. Hurwitz
Seybold UNIX in the Office Report
We were asked to look at Data General's new implementation of its UNIX
operating system and its overall UNIX strategy. We have evaluated Data
General's approach in terms of the sophistication of its UNIX
technology and its positioning in the marketplace.
SYNOPSIS OF IMPRESSIONS AND RECOMMENDATIONS
There has been a rapid acceleration of the movement towards UNIX over
the last year. This movement is expected to gain momentum because of
IBM's AIX announcements during the last six months. In addition, with
the Europeans, Japanese, and the U.S. Federal Government so strongly
behind UNIX and standards, most vendors are developing a
standards-based product strategy. Because competition will become even
more intense over the next several years, it is wise for Data General
to strive to present a strong and impressive UNIX strategy as quickly
OPERATING SYSTEM. Data General has done a creditable job with the
design and implementation of DG/UX version 4. It adheres to industry
standards while innovating in some key areas that are geared to the
commercialization of UNIX. These innovations are most notable in the
areas of I/O and support for symmetric multiprocessing. Kernel
redesign has made the operating system base modular and more efficient.
The strengths of DG/UX version 4 are:
1) Support for symmetric multiprocessing is very nicely designed and
should offer technical advantages for Data General over other
competitors in the marketplace. The key strengths include:
o The virtual processor model that effectively controls the use of
o Process management through the use of short-term and medium-term
schedulers. This two-level facility significantly adds to the
commercial effectiveness of DG's approach to symmetric
o Use of advanced scheduling algorithms, an added benefit in
helping to make this approach even more impressive.
o The interrupt mechanism, which is well designed for symmetric
multiprocessing because it is able to scale to the number of
processors and the number of devices being handled.
o The use of virtual-processor technology. This has important
implications for special-purpose server processes (a key
direction in the future of DNC).
o The handling of I/O, especially significant in the large scale
end-user environment. That all I/O is not handled by a single
CPU is a key advantage in the multiprocessor system.
2) The way that files are organized and safeguarded is an important
factor for commercial applications of DG/UX. This includes:
o Design of a logical file system to allow files (and file systems)
larger than the size of the physical media.
o Replication of the system database as well as the foundation for
disk mirroring to provide a high level of reliability and
backup. This is a feature that will be of particular interest to
commercial data processing organizations.
3) The modular design of the kernel and operating system itself will
position Data General as a creditable competitor. Key benefits of the
modular design include:
o Isolation of hardware dependencies.
o Ability to pull out parts of the operating system, such as the
virtual memory manager, and replace it with new technology that
o The streamlined design of the reconstructed kernel, which should
greatly improve performance over that of previous DG/UX operating
o New modularity which will allow DG flexibility to implement this
operating system on new hardware platforms with relative ease.
Therefore, scalability is a key feature (although the announced
direction for IBM's AIX). That the same operating system can be
implemented on different-sized processors and different
architectures is a critical success factor.
o Conforming to a Binary Compatibility Standard for the 88K
processors. This makes good marketing sense. In the long term,
an Applications Binary Interface would be advisable.
We expect that Hewlett-Packard will come up with an approach to
symmetric multiprocessing although to date, no such capability is
available. Therefore, this lead in technology will be limited to
between one and two years. In addition, while Digital Equipment
Corporation's symmetric multiprocessing software runs only on VMS and
does not perform at the same level of sophistication, DG must assume
that DEC will continue to improve its technology in this key area and
we expect that they will move it to UNIX.
DG needs to distinguish its approach to Virtual Memory Management from
those used by both Hewlett-Packard and IBM. Both HP/UX and AIX allow
for users to have transparent access to programs that do not fit into
physical memory by swapping segments in and out of main memory. DG's
approach allows users to have transparent access to programs that do
not fit into physical memory by demand paging on a per page basis. DG
needs to be clear on how its approach is superior to this alternative.
We believe there are several areas not included in the Release 4 that
should be made available either immediately or in the near future:
o National Language Support (NLS). We recommend that DG implement
Hewlett-Packard's NLS, which is viewed as a de facto standard.
o Support for real-time computing. This is perceived as an
increasingly important requirement in computing. Traditional
UNIX lacks this. More vendors are beginning to implement real
time in UNIX operating systems. Hewlett-Packard has included
extensive real-time operations such as assigning real-time
priorities to applications over lower priority processes and
time-based scheduling. In fact, HP's real time capabilities
sound similar to some of the scheduling and virtual process
control of the symmetric multiprocessing that DG is
implementing. For example, this is one of the design goals of
AT&T's System V.4. Although DG has the underpinnings for real
time, it needs to inform customers about these capabilities.
o User Interface. While we understand that X-Windows will be the
underlying networked windowing for DG's UNIX, the company needs
to come up with a standard user interface ("look and feel") as
well as tool kit. We recommend that DG hold off on making a
decision on user interface. We expect that X/Open will name a
user interface standard in October 1989. DG should comply with
their selection. DG should also consider implementing at least
the look of this interface on all its platforms. This is the
approach that DEC has already taken with its user interface.
APPLICATIONS ENVIRONMENT. The move towards standards will pose a
challenge for what we call the second-tier vendors. They are companies
such as Wang, Data General, NCR, Altos, and Hewlett-Packard - to name a
few. Within this grouping, each vendor must distinguish itself from
The Distinguishing Factor. What does it take? First, in terms of the
UNIX marketplace, it requires excellence in terms of the operating
system itself, service, support and an awareness that the company is
indeed at the leading edge of technology. And once a vendor adheres to
these standards, it will face an even bigger challenge - to distinguish
itself in the applications software arena. This will be a requirement
for survival within the next three to five years. It will happen once
there is what we call a level playing field. In other words, in the
long run, all vendors surviving will offer the same level of operating
system sophistication, the same amount of hardware power and at least
reasonable service and support. So what will be the distinguishing
feature at this point? Software! Office and communications software
will be what vendors will use to attract and hold onto their customer
base within five years. Such software will have to be superbly
designed and executed.
As much as possible, DG should offer the same applications on both the
proprietary and standard platform. This will give users freedom to
select their preferred environment. Also, it will allow users to
migrate to "standards" without forfeiting the applications they need.
From a strategic view it will prevent DG from appearing to be two
companies with two independent strategies. Providing SMTP facilities
so that UNIX and CEO system can exchange mail is a move in the right
Strengths of DG's approach to applications include:
o Encouraging key applications to be ported to DG/UX, especially in
the database arena. This makes good sense.
o Migrating DG's proprietary Infos DBMS to UNIX. This is a wise
o Offering industry standard ANSI SQL is a requirement that DG is
o Providing assistance to third parties to that they may port their
applications to DG hardware is a requirement.
Our perception is that Data General intends to have a leading edge
version of the UNIX operating system for both its proprietary hardware
(MV family) and the forthcoming Motorola 88000 series of RISC
processors. In recognition of the importance of the Intel 80386, Data
General intends to have offerings in the market segment, geared
primarily to the low end systems market. Therefore, Data General will
have a scalable approach to the marketplace, ranging from 8086 PCs to
large scale multiprocessor 88000 and MV systems.
This strategy is workable if the Motorola 88000 family does indeed
become one of the industry standard platforms. The 88000 is attractive
because it will be one of the industry standard RISC chips. In
addition, the 88000 will provide the type of scalability that DG
needs. DG should be aware that other RISC chips will also emerge as
important standards. For example, Intel has already announced a RISC
chip -- the 80960. While it is initially aimed at the military
marketplace, it will make its way into the commercial arena by 1989.
In addition, the full impact of Sun's SPARC chip has not been fully
realized. In brief, the 88000 will not the the only game in town.
While it is wise for DG to use the 88000 family as its standard
platform for the future, it should remain flexible, and recognize the
potential of another technology that may sometime in the future prove
even more promising. Simply put, it is best to have an open mind, and
the portability of DG/UX allows for this.
Keeping the proprietary MV hardware available also makes sense in terms
of protecting the company's existing customer base. Data General has
earned its reputation for reliable hardware. it should continue to
support low-end PCs (Intel 8086) as well as the intel 803886. the
80386 will become increasingly important as OS/2 emerges as a low end
solution for multitasking applications (both on standalones and in
local area networks). Over time, the 80386 will become the de facto
desktop machine. Therefore it is appropriate that DG implement both
multiuser and server products based on the Intel 80386.
Another important role for these processors such as the 80386 is as
servers for local area networks. DG should support these servers in
its overall strategy. For example, DG would be wise to support LAN
Manager both in the OS/2 arena and the LM/X (the HP/Microsoft UNIX
version of LAN Manager). With the power of DG/UX to support
specialized servers, the support for low-end servers should be a
fundamental part of its overall strategy.
APPROACH TO STANDARDS. With the announcement of a high quality UNIX
operating system with many advanced features, it is important that Data
General take steps to make its presence felt in the standards arena.
If DG is too cautious it will miss an opportunity. DG must be bold in
its approach to the UNIX standards environment. We were surprised to
learn how much effort DG has been putting into UNIX over the past
several years. I believe that the rest of the industry shares this
surprise. Focusing attention on some of the areas where DG has superb
UNIX technology, DG could establish itself as a first class UNIX