| >>>(mail forwarding headers removed)
I N T E R O F F I C E M E M O R A N D U M
Date: 04-May-1993 02:04pm EDT
From: Peter Lowber
LOWBER.PETER
Dept: U.S. Competitive Team
Tel No: 297-6341
TO: See Below
Subject: Gartner Mid-Range Conference Notes
I N T E R O F F I C E M E M O R A N D U M
Date: 03-May-1993 06:08pm EDT
From: BREUNIG
BREUNIG@VMSDEV@MRGATE@CSTEAM@MRO
Dept:
Tel No:
TO: PETER LOWBER@MRO
Subject: NOTES FROM 1993 GARTNER MIDRANGE CONFERENCE
+---------------------------+TM
| | | | | | | |
| d | i | g | i | t | a | l | I N T E R O F F I C E M E M O
| | | | | | | |
+---------------------------+
TO: Distribution DATE: May 3, 1993
FROM: Kevin Breunig/
Peter Lowber
CC: DEPT: OpenVMS Product Mgmt/U.S.
Competitive Sales Team
DTN: 381-1917/297-6341
LOC: ZKO3 / MRO1
NET: VMSDEV::BREUNIG/
CSTEAM::LOWBER
***THIS MEMO IS FROM KEVIN BREUNIG AND PETER LOWBER***
*** DIGITAL CONFIDENTIAL ***
SUBJECT: NOTES FROM GARTNER GROUP MIDRANGE CONFERENCE, APRIL 15-16, 1993
Due to the wealth of information at this conference and its importance to
Digital, we have combined our notes to highlight key points made at the Gartner
Group Midrange Conference.
HIGHLIGHTS (AND LOWLIGHTS):
o Gartner sent somewhat mixed messages about Digital. Digital was
represented as technology leader in key areas (clusters, RTR, AXP
performance) but struggling with sales force and channels issues. With
Lucente's appointment to VP of Sales and Marketing, Gartner raised its
probability that DEC will survive from 60% to 70%. Bottom line was
something to the effect of, the next year will be tough for Digital
customers, but if you see it through, it will be worth the wait (my
paraphrasing.)
o Despite this uncertainty, Gartner expects Digital to increase its share
of midrange systems revenues from 15% in 1993 to 18% in 1998. HP and
NCR are also expected to gain share, while IBM is expected to lose share
(22% --> 20% in the same period). However, Digital will have a "difficult
transition" as it is losing share today.
o Gartner presented clear timelines and futures for each Sun, HP and
IBM platform. For Digital, it focused almost exclusively on organizational
issues and previously announced Alpha products. I do not know if this is
because the organizational issues are foremost on customers' minds, or we
have failed to present Gartner with a vision for the future of our products.
o Fairly negative coverage on our NAS strategy (NAS diagram presented with
a large circle and slash superimposed) and TP strategy (described as
"confused.")
o Positive coverage on Digital's client/server infrastructure (Accessworks,
RTR, ACMS), although considered "proprietary". Reports from clients on
porting applications to Alpha very positive ("no reports of 'bad ports'").
o Choice quotes from Wes Melling - we need to turn around some of these
perceptions:
"DEC is culturally unable to write a positioning statement on
these [TP and client/server] products"
"You will see tremendous confusion in the field over the next
12 months..."
"I would not be surprised to see Rdb go away in two years."
"Digital rewarded the designer of its confusing, 5 API TP
strategy by making him VP of Software" (paraphrase).
"Alpha is not enough by itself to make Digital profitable."
o ON DEC's UNIX Strategies:
- DEC must have a strong UNIX and OSF/1 provides that.
Cultural issue: Can you find anyone in DEC who has their heart
in UNIX? Not Many.
- Digital believes strategically in a unified Unix strategy;
even if it does not believe in Unix emotionally, users should
not be concerned, as enough people understand the business
need for having a strong Unix offering.
o Midrange Evaluation Model (where OpenVMS historically leads) will not
be updated until July (this despite vendor submissions in January -
we should update with new AXP software availability and benchmark results).
o All vendors are talking about "clustering" -- it is important for Digital
to communicate its leadership in this area and force competitors to
compete using our definition of clusters -- which includes high degrees
of scalability and manageability.
o Server Price War:
- RADICAL Transformation in Mid-Range server pricing brought on in
1993 by Pentium Servers: Compaq, AST, et al-- 6-way, low-cost
servers. Commodity hw. Entry server prices around $10K. Pricing
models move from value-based to PC-Model.
- GG recommendation: Defer your choice for as long as possible until
the Mid-Range price war settles down.
o Tandem and DEC lead with hi-volume tp (NYSE), but this will be
overcome by 1996 when HP, RS6000 will have High-Availability,
Clustering with tp.
o Encina: powerful, with 3-level deployment and standards. A lot of
customers have bought into it.
o (Wes Melling) on UNIX: "I know of no UNIX System that has the
level of availability that I would feel comfortable with. HP's
Highest Level of Availability is on MPE and not UNIX. With
High-Availability UNIX Systems by 1996, I think UNIX will catch up."
o (Wes Melling) on AS/400: "Not for high-volume tp. I would
disqualify it for any situation for high-availability. I'm not
talking about a reliable box (AS400 very reliable box). AS/400 is
much like MVS: batch orientation: deliberate downtime for
re-ordering the data base, changing the software, etc."
o (Wes Melling) Why should we forget about OS/2?: ISv's are writing
for NT for Both Desktop and Server applications. By YE 93, NT will
have Repository, systems management tools, CORBA sw, tp monitor.
"NT will be light years ahead of OS/2 by YE 93. OS/2 will be
over-run."
KEYNOTE (MYRON KERSTETTER)
--------------------------
o Struggle will continue between end-users trying to build ad-hoc PC-LAN
based solutions, and MIS trying to impose control through architectures
and resource management. 65% probability that MIS will be able to
reassert control over IT spending as client/server management and
development tools improve to acceptable functionality.
o "Bottom up" IT spending (PCs, LANs, workgroup applications) will increase
from 35% of total IT spending today to 55% in 1996.
o Key direction setters for "bottom up" computing will be Microsoft,
Novell, and database, 4GL and tools ISVs. IBM, DEC, HP, Compaq, NetFRAME
and AST will be "along for the ride."
o "Top down" processing (batch, TP, PC integration), moving toward a
three tier model with fault and disaster tolerance, will continue to be
an important computing structure. Direction setters in "top down"
computing will be: IBM, DEC, HP, Oracle, Sybase.
o "We believe all the major mainframe vendors have resigned
themselves to replacement of their traditional mainframes by 1996"
o By 1994, midrange systems on a per-box basis will run at equal or
greater speeds than mainframes. Mainframes will continue to have some
throughput advantages, but at a 50-100% price/performance disadvantage.
o Disaster tolerant networked systems are expected to mature and supercede
hardware fault tolerant systems beginning in 1994. Gartner sees an
expansion of disaster tolerant capabilities:
CPU FAILOVER ---> STORAGE FAILOVER ---> SYSTEM FAILOVER ---> ENTERPRISE
FAILOVER
(Pair & Spare) (Mirroring, RAID) (Clustering) (Geographic
redundancy,
3-tier client/
server)
o Software architectures like SAA and NAS "failed to deliver on product
or industry leadership." Result is that customers will be taking a
pragmatic approach to building client/server solutions.
o Key requirements for "Enterprise Open Client/Server Processing":
SYSTEM MANAGEMENT
- Integrate system and network management
- Address bottom-up desktop as well as server and
traditional processing products
- Attract third party applications (populate
frameworks)
- Accommodate heterogeneous client/server environments
ONLINE TRANSACTION PROCESSING
- Focus sales/products vis a vis multiple OLTP models
- Client/server requirements in OLTP
- De facto standard API support
MIDDLEWARE
- Distributable applications (c/s and replicated)
- Interoperability
- Portability
- Global data access
SYSTEM/SERVER PERFORMANCE
- Absolute performance potential
- Scalability of performance
- Investment protection (upgradability)
o Midrange System and Server Market ($15K - $2M)
1993 1998
TOTAL MARKET $47B $60B ~4-5% CAGR
(NEW HARDWARE AND
SOFTWARE) WORLDWIDE
MARKET SHARE
IBM 22% 20%
DIGITAL 15% 18%
HP 13% 17%
NCR 6% 7%
UNIX 20%* 30%
NT -- 26%
*(50/50 split workstations and servers)
o Surviving midrange system and server architectures will be AS/400,
OpenVMS (VAX moving to Alpha) and Unix and NT.
CLIENT/SERVER TECHNOLOGY (PAUL MCGUCKIN)
---------------------------------------
o Gartner defines five styles of client/server computing:
STYLE BUILDING BLOCKS
Distributed Screen Scrapers (Easel, Mozart)
Presentation
Remote X-Windows
Presentation
Distributed Logic TP Monitor or 3GL RPC based applications
Remote Data Oracle, Sybase, Progress applications
Management Query tools for RDBMS
Distributed Complex user written applications using
Database RDBMS with two-phase commit
o Distributed presentation and remote presentation are considered to
be of limited use. The Remote Data Management style often used today
will give way to a 3-tier Distributed Logic model, which will require
key pieces of middleware, such as:
- message routing
- security
- directory services
- integrity and recovery (e.g. RTR, TP monitors, message queuing)
o Lack of a comprehensive client/server system management solution will
remain the biggest cost and reliability impediment to mission-critical
use of client/server. "You should not rely on systems vendors to solve
the heterogeneous management problem."
- Mid-Range players are not providing multi-vendor solutions
(includes HP Open View, which is UNIX-based, and DEC Polycenter).
- Opportunity: no one is bridging the three islands (including MF's,
and PC-LANS- NetWare). "No comprehensive set of tools.
o Islands of Middleware will exist at many customer sites:
- MAJOR disconnect between DCE-based in-house applications (which
many customers are planning on) and third party applications. Third
party dbms's have their own "rpc's" (not support DCE); NetWare 4.0
is in the same boat.
o The sophisticated middleware required for high volume client/server TP
will imply vendor lock in and high levels of programming expertise
until 1996; ability to code complex 3GL code with RPC and threads will
be required in short-term.
o Midrange price/performance expected to improve 35% annually through
1996; Compaq has announced $5,436 per TPC-A. $4000/TPC-A is probably
the theoretical lower limit.
o DCE is the preferred client/server infrastructure, according to Gartner
Group. Availability of DCE "secure core:"
IBM AIX 1H93
HP-UX 1H93
DEC OSF/1 1H93
DEC VMS 2H93
HP MPE/IX 2H93
SYSTEM V 2H93
NETWARE LATE 93 P = 0.7
NT LATE 93 P = 0.6
IBM OS/400 LATE 94
IBM MVS LATE 94
Development tools will emerge in late 1993, early 1994
o "Open Proprietary" operating systems (OpenVMS, MPE) will typically not
be chosen as basis for open client/server computing (p=0.6). According
to Gartner, "the allure of high-functionality middleware and programmer
inertia in using the familiar proprietary interfaces will often result in
users ending up with the same lock-in that they had before the advent of
"open proprietary" operating systems.
o Windows NT will emerge by 2H 94 as a credible if not functionality
leading Enterprise Server Platform (ESP). GG's evaluation shows that
NT will be ahead of other Mid-Range options for Middleware, and Desktop
Integration, based on the ESP model. It will be behind slightly in Network
Infrastructure; and behind in Systems Management. Significantly behind in
hardware availability.
- NT by 1995-96 will have a "steep" growth rate challenging UNIX.
- "Expect DEC to add a lot of value; potential competitive Advantage
for the Server.
- NCR is "holding at arms length" away from NT; walking away from
NT opportunities currently.
NT STRENGTHS (2H94) NT CHALLENGES
OLE Object Sharing Relative Immaturity
Hermes software distribution Comprehensive Network
and license management and Systems Management
ISV attention (server and
client) SMP learning curve
Bottom up buying preferences Top down buying perferences
o Forte was cited as one of a new breed of emerging development tools
enabling speed of development, RDBMS independence, and ability to
support a variety of target deployment systems and configurations.
o Pentium:
- at least 6 vendors will ship Pentium servers in May
- 2x performance of 486
- uniprocessor performance will approach midrange of RISC
architectures, but will not be competitive in floating point
performance
o Overall server/host market expected to grow less than two percent
per year. However, Unix will show tremendous growth (25-30% CAGR),
growing to $30B in revenues (HW and SW) in 1997. NT will also grow,
slowly between 1993 and 1994, then ramping up and reaching $10B in
revenues in 1997. NetWare expected to grow slightly, OS/400 and
OpenVMS expected to decline slightly through 1997. MVS sales will
decline from about $20B today to $15B in 1997.
NEW MODELS FOR TP (WES MELLING)
-------------------------------
o Promised a vendor comparison of TP architectures in 60 days
o Divided TP Market into three segments:
-- High volume/hypercritical (HV/HC): e.g. NYSE, SABRE; very high volume,
downtime very costly
-- Medium volume/mission-critical (MV/MC): medium volume, but high
availability and integrity are critical. Examples: hospital systems,
sales order entry systems. Cited L.L. Bean as an example, which
processes 400K order lines per day, but in a 24x7 environment.
-- Replicated site with roll-up (RSRU): volume not an issue, but
distributability and remote system management are important. Examples:
Home Depot, Wal-Mart, Hyatt, Blockbuster, Toys R Us.
o For less than 200 active users, a TP monitor is not necessary. For
1000+ users, it is critical. Traditional TP monitors are being broken
up into a set of distributed systems services.
o Recommended a server-centric (rather than client-centric or "fat client")
approach to high volume TP. Client-centric approaches tend to create
management problems and high volumes of message traffic.
o TP monitors are expected to converge on OSI-TP, OSI-CCR (Commit
Control and Recovery) and X/Open DTP APIs, with product support rolling
out in 2H94, and real heterogeneous interoperability beginning in 1996.
o Spent some time comparing vendor TP strategies:
IBM: 1 API everywhere (CICS); multiple underlying technologies
(S/390, HP, RS/6000, AS/400...); highly portable
NCR: 1 API and one underlying technology (TopEnd); supports
NCR 3000, HP 9000 and RS/6000
HP: multiple APIs (CICS, STDL, Encina) with one underlying
technology (Encina)
Digital: "most confused strategy"; 5 APIs (ACMS, CICS, MUMPS,
Tuxedo, RTR) with five underlying technologies. ACMS
and RTR are the most functional, but also the least
portable.
o Believes Digital and Tandem will maintain their lead in HV/HC systems
through 1996; slide had picture of Disaster Tolerant VAXcluster running
FDDI (no mention of MDF).
o For lower volume MV/MC applications, other platforms with clustering,
fault tolerant servers and 3-tier client/server capabilities become
acceptable. Examples: Encore, HP 3000, ES/9000, RS/6000, NCR, Pyramid.
o Replicated Site w/ Rollup needs can be met with a RDBMS with client/
server support (Oracle, Sybase, Informix etc.) or in some cases by
a Novell LAN with a database NLM.
o Quote: "I know of no Unix system [today] running high volume TP
in a 24x7 environment."
o AS/400 is disqualified from the MV/MC segment as he "has yet to
see a high availability AS/400 configuration." It is difficult to
work around planned downtime in AS/400 environment.
o "Shared nothing" systems will replace clusters over time, as any
shared resource becomes a potential bottleneck.
CLIENT/SERVER MANAGEMENT (PAUL MCGUCKIN)
----------------------------------------
o Client/server computing changes the requirements for system management:
NEW MANAGEMENT DISCIPLINES/TECHNOLOGIES:
> Software Distribution
> Network-based Backup/Restore
> License Management
> Middleware Management (e.g. Directory Services)
REDEFINED MANAGEMENT DISCIPLINES/TECHNOLOGIES:
> Security (authorization/authentication)
> Configuration and change management (e.g. software version control)
> Performance monitoring/capacity planning
o Maybe 8 management frameworks are being promoted by vendors today, and
the proliferation of frameworks will continue through 1996. An
integrated platform like Polycenter, OSF DME or Candle CT is preferable
to the "manager of managers" approach used by HP OpenView, UNMA or
NetView Service Point.
o DME Framework based management applications will not be available until
1995, and will not reach critical mass until 1996.
o OSF is expected to announce that IBM will be the DME integrator, and IBM's
SOM (System Object Model), which is CORBA compliant, is likely to replace
Tivoli's object broker. IBM's involvement increases the likelihood that
that DME will be available in 1995.
o DEC's Polycenter provides
- Integrated Platform of network and systems management
- single way of storing data
- recent enhancements allow for easier integration of third
party applications
- integrated with PC's; can launch Polycenter applications
from PC menu
o SNMP Mangers (like HP's Open View):
- provide a useful but superficial level of integration
- increasingly being positioned as systems management
(as well as network management) platforms
- SNMP is not suitable for systems management (low level of
security)
- OpenView is catching on
- Expect HP to announce in May a new OpenView to handle
network and systems management for HP, Sun, IBM RS6000
o Backup and Hierarchical Storage Management: Legato Networker
- (with Legato) Can Backup NetWare and UNIX servers
- one of few products that can b-u NetWare from a UNIX server
[NOTE: DECnsr is based on Legato]
o DMTF (Desktop Management Task Force) and DSIS (Distributed Support
Information Standards Group) will dominate standards setting for desktop
management. Novell and Microsoft will support DMTF drivers, and DMTF
will also support an SNMP interface for management consoles. Lack of
DME integration direction may slow acceptance in large companies.
o LAN based Electronic Software Distribution Systems will mature over the
next 18 months. Users should demand electronic software distribution from
their vendors. Major opportunity with no clear leader today. Unix vendors
may be making a mistake by focusing too much on the desktop.
o License Managers expected to converge on one de facto standard around
1995, and are expected to have 50% market share in that timeframe. IBM
is expected to focus its efforts around NetLS. Only Gradient's NetLS,
Digital's LMF, and Highland's FlexLm are flexible enough to accommodate
new licensing models. HP, IBM and the OSF are supporting NetLS.
o "Legacy" management vendors are expected to focus more on downsizing
market than c/s. CA is porting to Unix, Candle is porting tools to
AS/400 and Unix etc.
o Hermes provides a good foundation for management (e.g. license
management/metering, ease of use) but will not be available until 1H94.
(Note that CA has announced that Unicenter will be ported to NT).
o Given the choice between Pathworks and NT for managing Windows applications,
NT and Hermes "probably the better choice." (Response to audience
question).
IBM (MYRON KERSTETTER)
----------------------
AS/400
o IBM had a tough year in 1992. AS/400 revenues were down, RS/6000
revenues grew only 30% (below a planned 50% growth rate). Mainframe
revenues are expected to continue to decline 10-15% per year through
1995. 1992 revenue mix:
- ES/9000: $35B
- AS/400: $14-15B
- RS/6000: $2B
o Less and less reason and incentive to create Common sw between IBM
platforms: SAA Fallout; The new IBM is the reversal of SAA
o Single Sales Force: the Flaw of the New IBM
- EXPECT IBM to separate the Sales Force
[NOTE: According to Open System Today, this is beginning to
happen as IBM is training a specialized RS6000 sales force]
- The Sales Force now has Product Specialists
- The IBM General Business Org
--- in Canada now- focused on third parties; Partnership
Management; for RS6000, AS400, PS2
--- this could emerge as a major structural change
o ES9000: Low-Cost CMOS (1994-95)
- Gartner believes IBM will not succeed in shoring up ES9000
revenues
- Continued Erosion of MF's
o IBM looking at developing a common RISC architecture for AS/400,
RS/6000, ES/9000 and OS/2, possibly based on PowerPC with different
"personalities" for each OS.
o IBM will be a catch-up Follower and Unable to return to
architectural leadership
- Top Down SAA Failed
- Industry Moving Bottom Up
- IBM Open C/S: "Open Distributed Computing": No Leadership;
o AS/400
- While IBM is committed to multi-vendor support for AS400
(networking and PC clients), the AS/400 is rated low by
Gartner in PC usability factors (solutions compatible with PC
applications and network-based systems management) and
Platform Optimization (AS400 optimized for 5250 timesharing
only)
- LAN performance is poor for AS400; need better I/O
bandwidth
- IBM is positioning the AS400 for C/S Remote Data Access
(with third parties like Gupta and with DRDA)
---[NOTE: BUT remember the AS400 Cannot Support third party
DBMS's like Oracle, Sybase, etc. This SHOULD eliminate it
from most Information Access Applications in Enterprise
Client/Server--- Peter's opinion].
o CICS will not be the preferred API for TP applications; developers
will continue to use native OS/400 interfaces
-- NO ONE (no customer) at the Gartner AS400 user session
(about 35 AS400 users) said they were interested in CICS/400
- OLTP Strategy: AS/400 is Low-Volume; not 7X24X52
-- AS400 apps use OS/400, not tp monitor
o AS/400 offers online backup today, but it causes degraded performance.
Loss of a CPU in an SMP configuration crashes entire system.
o AS/400 will continue to be a good general purpose application server,
but a relatively weak database server.
o In answering a Question about AS400 High-Availability Myron said
that "it can take days to recover files if a disk drive crashes
[because of single-level store]; there is no clustering; no fault
tolerance.
o AS/400 Futures:
Message Q (IBM developed this with SSI- not shipping yet)
is new and is IBM-centric
Stored Procedures (2H93 - V2R3)
AppleTalk, SPX/IPX support (2H93)
Port of Candle system management tools (IBM will have exclusive
rights to selling Candle/400 products) (year end 1993)
G Models w/ I/O RISC processor, upgradeable (1H94)
POSIX 1003.1, .4, .4a, DCE, XPG4 "enabled" (late 1994)
H Models - minor refresh of line (1995)
Delivery of Object Model (SOM) for AS400 (1995-6)
Plans to support third party databases? NONE
"AS4000" RISC; 64-bit using PowerPC 620 chip- "expect easy migration"
(1996)
[it should be noted that Gartner is no longer listing "clustering"
as an AS/400 future - RS/6000 will fill this role.]
o Gartner projects continued unit growth for AS/400 within its installed
base [Note: AS400 revenue declined by 2% last year; -10-15% in Europe;
Decline in 1Q 93 likely since IBM said nothing about AS400 in
its 1Q report]
o AS/400 TPC-A results are not competitive, even with F-series announcement.
TPC-C results are better, but IBM will continue to price more aggressively'
at low end than the high end:
SYSTEM tpm K$/tpm
AS/400 F80 580.3 $3.28
AS/400 F35 98.6 $2.44
AS/400 F10 66.1 $2.10
RS/6000 570 356.4 $2.00
HP9000 H40 406.6 $2.77
o AS/400 vs. OpenVMS and Unix
AS/400 VAX/VMS UNIX
"OFF THE SHELF" COMMERCIAL E F F
APPLICATIONS
COST OF STAFF E G G
COST OF SYSTEM G G E
HIGH PERFORMANCE OLTP G E G
PC INTEGRATION, CLIENT/SERVER G G G
ARCHITECTURAL LONGEVITY G F* G
HIGH AVAILABILITY G E **
SYSTEMS MANAGEMENT F G P++
PERFORMANCE SCALIABILITY F E G
ARCHITECTURAL ADAPTABILITY F E G
OPEN SYSTEMS COMPLIANCE F G G
COMPUTE INTENSIVE P G E
KEY: E = EXCELLENT, G = GOOD, F = FAIR, P = POOR
* Digital is shifting from VAX to AXP; OpenVMS functionality should be fully
ported to AXP by mid-1994
** HA features vary greatly by Unix version
++ Unix system management improving rapidly, with capabilities varying by
Unix version
RS/6000
o has't turned the corner; still maturing; many IBM customers have
RS6000 pilots; IBM has not gained much in
- the oposite of the AS400 profile: platform optmization
strong, but solution level integration weak (reliance on third
parties for this); PC usability is also weak
- Will IBM's added value succeed?? Not clear that customers
want CICS, DRDA, DB2/6000 for RS6000
- Focus on Systems/network management: NetView/6000 is
based on HP OpenView: point solution; lacks a framework
for integration; IBM and HP are working on this; likely to
become the COSE std
o RS/6000 continues to lag in Unix performance and Scalability; SMP
not expected to be available until Q1CY94, when RIOS2 systems are
available. RIOS2 systems are expected to approach 250 tpsA.
o HACMP "clusters" now support failover between two nodes; however,
special IBM disk drives and Oracle V7 are required, making it a
"closed" solution. Expected to be enhanced to support 4 nodes in future.
o RS/6000 & AIX Futures
CICS/6000 (June 1993)
SMP systems (2-4 way), 32-bit, RIOS2 based (Q1CY94)
AIX V4.1: supports SMP, SOM, DCE Distributed File System,
DB2/6000, Native NetWare (Q1CY94)
RIOS-2
-- stop gap; nine chip Module; 32-bit; YE93
-- to cover the high-end not covered by PowerPC 601,602,604
-- 240TPS (Alpha is faster now)
PowerPC 620: YE94; ship 1H95
64-bit; MACH-3 kernel operating system
- Shared Memory Cluster (SMC): 1H95
--- this is very confusing; SMC requires some changes in
applications to implement; why do this when you have SMP?
o IBM will have similar problems to Digital in positioning AS/400 and
RS/6000. Positioning will increasingly be done by customers, many of
whom have both.
HEWLETT-PACKARD (GEORGE WEISS)
------------------------------
o HP background
- now at 92,000 employees
- FY1992 revenue mix:
TOTAL REVENUES $16.4B
NON-COMPUTER RELATED $ 4.3B
COMPUTER RELATED $12.1B
STANDARD PRODUCTS $ 5.7B
(PCs, Printers, disks)
PROFESSIONAL SERVICES $ 0.6B
SYSTEMS $ 5.8B
HP9000 76%
HP3000 24%
- Installed base:
HP3000 (MPE) 55,000*
HP9000 (UNIX) 30,000
*NOTE: only 1/3 of HP3000 base is RISC based MPE/iX
Allbase runs on 20% of HP3000 systems
Only 85 high end HP3000 Corporate Business Systems installed
Total Corporate Business Systems orders = 200 orders,
50/50 split MPE/HP-UX
o HP's goal is to raise its credibility with Fortune 500 class accounts
on par with IBM and Digital. HP is expected to grow from being a
"strategic vendor" in 20% of F500 accounts today to being "strategic"
in 60% of F500 accounts by 1995
o Key Issue: How to make the transition to an Enterprise Vendor
- George asked the HP users in the audience if they felt HP
was a strategic enterprise vendor and the vote overwhelmingly
was YES. This was not the case at past GG MidRange
Conferences
o George clearly sees DEC as the major competitor for HP
o Challenge: Build a strong MIS relationship and Professional
Services ability to build an infrastructure is impacted by the HW
Price Wars
o HP's Primary Strategy is HW Sales
o Trying to provide MF Alternative Systems: 100 S/890's shipped so
far (50 MPE; 50 UX) replacing IBM 43XX's MF's
o Offer Open Systems Foundation Building Blocks
- HP proprietary Sw (Open View, Performance View, SwitchOver)
with hooks for third party solutions
o HP is targeting AS400's (but this is a shaky strategy: How many
AS400 users will be willing to move to HP/UX?
o HP SharePlex: Not VAXclusters; not large clusters for scaling
performance, But can balance the workload; remote shadowing files
over WANS for disaster recovery
o UNIX not reliable for hi-volume OLTP
- working on kernel UX since 1981; stable kernel; keep building on
this to create reliable environment; OSF/1 would have been an
interruption; minimize the transition disruptions; 64-bit is
something you can wait for; next release of UX 10.0 offers threads
[available now with OSF/1]
- Question: Given the importanc eof the Kernel, why did HP stay with
SVR3 instaed of SVR4?
- George gave the same answer: THat HP will only make one
transition- to the micro-kernel technology. Making a transition now
would "only cause a hiccup in sales"
o OpenView----> working with Novell for OpenView with NetWare
- offer CA Unicenter as well (as an option)
o Client/Server: to offer pre-configured solutions with UX Servers
- HP Professional Services Organization ($600M) is adding people
with "the intent of providing more of a systems integration."
HP is moving gradually and may vary by account in terms of their
ability.
o HP expected to enter superserver business with Pentium based SMP systems
running NT and NetWare 4.0 (2H93)
o Current Corporate Business System performance (HP3000-99x and HP9000-89x)
at 180-200 tpsA (uniprocessor) and 380 tpsA (4-way SMP) (all figures are
host based).
o HP is expected to announce second generation of Corporate Business Systems
in Q3 or Q4 for CY93 with 8-way SMP, and performance of 230 tpsA (uni) to
1200 tpsA (8-way). Pricing is expected to be in the $8000/tpsA range.
o GG believes HP can boost the performance of PA-RISC up to five times
today's performance, with 200-250 Mhz clock speeds, in five years.
o HP will continue to invest in MPE/iX to improve its robustness relative
to MVS as part of its Mainframe Alternative Program. However, there is
only a 50% probability of significant investment beyond 1995. PA-RISC
is expected to maintain HP3000 longevity through the end of the decade,
although HP3000 system revenues are expected to decline from $1.5B in
1992 to about $1 billion annually.
o Investments in MPE/iX robustness/availability include: subsystem file dumps,
page de-allocation, predicative failure reporting, job scheduling, resource
management, extensions to SharePlex "clusters".
o HP3000 Futures:
SharePlex Cluster/Disaster Recovery capabilities (1H93)
SharePlex Systems Management (later in 1H93)
8-way SMP, RAID 5, Encina and CICS TP (2H93)
2-way Midrange systems, FDDI support in SharePlex (1H94)
Improved OS resilience & recovery, job scheduling,
resource management (1H94)
Mixed HP-UX and MPE SharePlex (2H94)
4-way 120 Mhz Midrange (2H94)
Third Party OpenView Management Applications - CA, Candle,
Legent (2H94)
16-way SMP (1995)
o HP will continue to invest in HP-UX through 1995, with gradual addition
of OSF capabilities. For example, OSF commands and libraries and DCE
are expected in the HP-UX V10.0 (2H93)
o HP will also continue to add high availability features to HP-UX. Also
expected in HP-UX V10.0: transparent switchover, distributed lock
manager, fast file recovery, workload priority scheduling, striping,
disaster recovery.
o HP9000 Futures:
OpenView System Management, threads, fast file recovery
(2H93 - HP-UX 10.0)
8-way SMP on 890 series, using 7100 chip,
up to 1200 tpsA (2H93)
RAID5 (2H93)
Encina, CICS TP; DCE Services, OSF Commands & Libs (2H93)
2-way Midrange Systems (2H93)
Mixed HP-UX, MPE/iX SharePlex (1H94)
Improved OS Robustness, recovery, switchover, workload mgmt
(2H94 - HP-UX 11.0)
4-way SMP 200 Mhz Midrange (2H94)
SharePlex Disaster Recovery (1995)
16-way SMP (1995)
New Microkernel based OS (OSF based) (1996)
o HP will adapt OpenView for systems management across MPE and UX by 1996
SUN MICROSYSTEMS (GEORGE WEISS)
-------------------------------
o Background:
- FY92 revenues (FY ending June 92): $3.6B ($1.5 commercial
applications)
- FY92 employees: 12,800 ==> revenue/employee @ $280K
- FY92 channels mix:
DIRECT 46%
VAR/DISTRIBUTOR 34%
OEM 20%
- CY1992 workstation/workstation server market share by revenue
(total revenue = $10.8 billion)
SUN 32%
HP/APOLLO 18%
DEC 10%
IBM 16%
SGI 8%
OTHER (INCLUDES
CLONES) 16%
- Estimated mix of workstation and workstation server products:
> 50% low end
> 35% midrange
> 15% high end
- workstation and workstation server installed base
> SPARC: 780,000 (I believe this includes clones)
> IBM POWER: 120,000
> HP PA-RISC: 135,000
o Sun's strategy is primarily volume-driven; Gartner says Sun has the goal
of being the "Intel of the Unix market." Sun will focus on systems
downsizing opportunities, stressing price/performance. Sun believes
it has a superior business model to HP, DEC and IBM with a much lower
cost structure and the ability to respond more quickly to market changes.
This high-leverage strategy carries some risk: for example, TI has
had problems supplying high end SuperSPARC (Viking) chips.
A chart was provided showing:
LEVERAGE + PRICE + BALANCED PERFORMANCE ==> VOLUME
(Channels, OS, Mfg, Appls,
Semiconductor Foudries,
Binary Specifications)
o Sun will have problems in the downsizing market as a) it still needs to
forge close ties to MIS, and b) it lacks key enterprise server capabilities.
Missing components include:
- Large Storage Management
- Data Integrity
- Systems Management
- High availability
- Data warehousing/extraction
- Open applications development and APIs
- PC LAN/workgroup development tools
- High performance OLTP
- Leveraging existing middleware (ONC, NFS, NIS+ etc.)
o Product futures:
High Availability Failover/Clusters (3Q93)
[this will be similar to limited failover in IBM's HACMP i.e.
loosely coupled systems running Oracle V7]
Software License Management (Q493)
XPG4 Base Branding (Q493)
DCE Support Package (CDS, RPC, DTS, Security) (Q194)
ONC+ (1H94 - includes 64-bit file support)
System Management Framework (Q294)
Federated Services (1995)
CICS API, CA tools (1995)
o SPARCcenter 1000 expected announcement soon; 4-Way SMP @$45K entry
o SPARCcenter 2000 expected to scale to 325-350 tpsA (4-way) by mid-1993;
may reach 700 tpsA by 1H94
o Perhaps in response to AXP performance, Sun is now saying it can
get by with 10-20% less raw performance than other RISC implementations
if it can offer products at 40-50% less cost. Sun is showing a roadmap
for its next generation RISC chip (UltraSPARC), which will begin rolling
out in 1995. It is expected to run at 140-500 Mhz, scale to over 20-way
SMP, and initially run at 300 SPECfp, scaling to over 1800 SPECfp in
1997-8.
NCR (PAUL MCGUCKIN)
-------------------
* Severe product delays have hurt NCR credibility and there's
customer skepticism. ISV's have moved their attention away form NCR
to HP, Sun, IBM.
* NCR has lost market share over the last two years. But now they
do have the products to ship
* Within the last month, NCR consolidated marketing to One Group
* Close development relationship with Intel for Pentium. NCR will
announce Pentium end-May. BUT it Won't support NT! NCR has a
passive position on NT as a server. NT will run as a client, but
NCR has walked away from NT opportunities this year! NCR's
strategy is SVR4-based.
Products
NCR 3450, 4X486 228 TPS $6,302 (2nd best behind Compaq)
(Oracle 3430
DIGITAL: WILL A VISION EMERGE? (WES MELLING)
--------------------------------------------
o DEC is trying to do it all. What's the new Vision statement from
the CBU's? (Drafting a vision statement is one of Lucente's top
priorities).
o Commodity Business:
- DEC's PC and Disk Businesses growing Fast
- DEC has re-organized the Business here; not just product
o Wes was positive on Alpha, which "on a box for box basis is
running twice as fast as HP [9000] and IBM [RS/6000] boxes." He did
not discuss any Alpha or software futures. These performance
numbers will allow Digital to "reenter the technical computing
market in strength."
o Alpha:
- 4,000+ installed, but not Production yet
- We hear the hw runs just like a VAX
-- OpenVMS runs "as good as any first version sw we have
ever seen on a new platform.. performance is excellent"
- Porting from OpenVMS VAX to OpenVMS AXP is easy
- "We havent heard of a single bad port"
-- Medusa: 3 level rotational graphics; 250K lines C code--
3 man weeks to port--- complex applic-- quick port
- COBOL port easy; ran 16X faster than on VAX!!!
* 550K VAXes installed so Alpha will succeed simply selling to
installed base
- and DEC will get back into the engineering market in
strength with Alpha
* BUT how does DEC succeed beyond installed base???
[NOTE: Wes nowhere evaluated DEC's OSF/1 AXP, and only referred to
DEC's UNIX capability in response to a question]
o However, Alpha alone will not ensure Digital's success. He is concerned
that "fast iron" will lull Digital into complacency, and divert attention
from urgent changes needed in marketing, sales, channels etc.
o Most likely scenario (p=0.6) is that the Alpha PC succeeds in getting
only 1-2% of the PC market, with a focus on high-end servers. Assumes
Windows NT is slow to penetrate the PC market because of system resource
requirements.
o Was amazed that Digital released its first TPC benchmark for Alpha on
a DEC 10000 - Gartner had forecast $7300 per tpsA, and Digital came
in at $8300. However, AXP is expected to be a price/performance leader.
o Digital is backing away from NAS, and will not port most components
to other platforms. TP strategy is confused, with five different APIs
supported.
o DEC's Richness of offering for Client/Server
- RTR, ACMS, Accessworks, Polycenter, MessageQ, ACA Services,
Pathworks
- BUT DEC is not able to position its client/server offerings
- And the more functional DEC gets, the more proprietary
- Trade off: Advanced Exclusive Functionality vs Equivalent
Functionality (open, portable) with Leadership Price/perf
(Alpha)
o Based on Lucente appointment, raised probability of Digital's
survival from 60 to 70 percent. Survival = ability to build new
commodity channels and business models, and compete effectively at
all levels of integration. Noted that this is the first time
Digital has put someone in charge of sales that has sales
experience. "You are witnessing one of the biggest corporate
cultural changes at Digital that you will ever see."
o DEC's cluster technology: 1985 VAXcluster is comparable to today's
clustering from some competitors.
- HP et al do not provide automatic load balancing, automatic
reconfiguration, logical view of applications by multiple
servers
o In response to audience questions:
- on DEC's UNIX Strategies:
- DEC must have a strong UNIX and OSF/1 provides that.
Digital is committed to a unified Unix strategy.
Cultural issue:
- Can you find anyone in DEC who has their heart in UNIX?
Not Many.
- Digital believes strategically in a unified Unix strategy;
even if it does not believe in Unix emotionally, users should
not be concerned, as enough people understand the business
need for having a strong Unix offering.
- NT will be functionally equivalent to OpenVMS in 1996 (70%
probability. However, OpenVMS is still the "best data center solution"
for the next three years.
- Does not believe Digital is making money with Rdb, which is increasingly
being focused on high volume TP applications. Said he would not
be surprised to see Rdb "go away" in two years.
- Digital's exclusion/non-participation on COSE is really a non-issue.
Digital already supports most of the COSE standards, and has no
motivation to beat up on Microsoft.
DIGITAL USER BREAKOUT SESSION
-----------------------------
o (Kevin) also sat in on the Digital user breakout. There was a lot of concern
about sales and channels issues. A few customers were very vocal; many
of the customers just listened. Representative customer comments:
- sales force is confused about channels policy. Account team does the
leg work to set up the sale, and then Pioneer or Avnet are brought in
to take the order. Pioneer or Avnet end up selling the product and
bundling in a service contract.
- sales force demoralized, don't know what Digital's vision is, don't
know where next cuts will take place
- only a few, usually the best, people are left in sales support. Many
former pre-sales consultants are now chargeable. Sales support people
now seem to supporting more customers, always on a plane, etc.
- no corporate vision around client/server (GE Corporate)
- problems with All-in-1 support, CSC would not resolve, had to
elevate (Corning)
>>>(distribution list of approximately 300 people removed)
|
| Digital's strategic relational database management system, DEC Rdb,
has recently received some extremely negative comments from portions
of the consulting community. We would like to take this opportunity to
refute some of the rash and uninformed statements that you may have
heard, and assure the you of Digital's continuing commitment to Rdb.
With regard to a particular comment about the impending demise of Rdb
within the next two years, we would like to assert that Digital is
fully committed to support and enhance Rdb for out installed base of
over 30,000 installations, and to drive forward with a strategy to
support new directions in the IT market. Rdb is clearly poised for a
new and significant period in its life cycle -- a period that will
demonstrate Rdb's leadership in performance, integration, and
innovation for the DBMS market. The vision and strategy for this
exciting future is described below.
Over the past year we have increased our investments in research,
development, and capital support for Rdb to meet the needs of our
customer base. This investment has made possible world record
performance, highly tunable systems, robust data integrity and
security, and delivery of leading edge standards and distributed
capabilities. We have further expanded and enhanced our support
organization, to become more responsive to the needs of today's
changing market. The market today demands outstanding performance,
interoperability, and portability, and Rdb is moving to provide the
best-in-class solution meet these needs.
We have announced world record uniprocessor performance for a relational
DBMS using the TPC-A benchmark, record sort performance using the
Datamation Sort benchmark on Alpha (an Advanced Development project),
and directions in support of VLDB. We have announced directions that
include the porting of Rdb to the OSF/1 and the Windows NT operating
systems on AXP systems -- synchronized with enhancement of our OpenVMS
offerings. We have received awards and recognitions from Datamation,
Computerworld, and Digital News and Review. We view these achievements,
awards, and directions as the start of a significant period in Rdb's
life cycle, not the end.
Digital's database strategy is driven by the direction of the market,
and the competitive environment. As a company, Digital sees three
significant trends in the IT market:
o Downsizing - the trend away from mainframe datacenters to fast,
economical, desktop PCs and workstations.
o Client/server computing - the integration of these fast and
economical desktop machines into powerful and coherent distributed
systems.
o Very Large Databases (VLDBs) - the growth of databases up to
hundreds and even thousands of gigabytes in size.
We view these trends as complementary, not divergent. All of these
trends present significant opportunities for Digital systems,
software, and services -- particularly around the new AXP systems as
the focus for downsizing and client/server applications.
While we believe that Rdb and AXP systems provide the best solutions to
your database management needs, we recognize that there are certain key
vendors in the market who have widespread acceptance and demand from
existing and potential customers. These vendors have significant
presence and customer loyalty, both in existing Digital customer
installations, and in IS shops where we would like to develop a
presence.
The database strategy builds on these assumptions with three thrusts:
1) Provide the world's best database performance, with leadership
production quality relational DBMS and an emphasis on graceful
handling of VLDBs, satisfying the needs of downsizing and
client/server applications;
2) Provide world-class data integration across multiple relational
database managers, from multiple vendors and on multiple hardware
platforms;
3) Make Digital's AXP systems the hardware platform of choice for
database management, for any vendor's DBMS.
Digital's investment and support of Rdb and related interoperability
products is the result of the first two elements in the strategy.
Digital's relationships with Oracle, Sybase, Progress, IBI, and other
RDBMS vendors are a direct result of the final element in the
strategy.
In conclusion, we see that our existing customers, and the companies
that we want to become our customers, are deeply committed to multiple
product offerings. We also realize that a spirit of fierce competition
without limiting choices best suits the reality of complex IT
environments. Our strategy and focus is on delivering best data
manager, Rdb, on open platforms (OpenVMS, OSF, and NT), adhering to open
standards, such as SQL and XA.
Further, we will deliver the best data interoperability through
announced and planned RdbAccess and database integration products. We
believe that your investments in other data managers can be preserved,
while the best choice for new or re- engineered systems will obviously
be Rdb.
We respectfully believe that independent consultants who have
extrapolated the potential demise of Rdb are wrong, and we have no
intention of canceling Rdb or even of reducing our investment in it.
We further view comments by third parties about Digital's database
plans as marketing ploys -- spreading fear, uncertainty, and doubt,
without any basis in fact. While the avowed philosophy of at least one
independent database vendor is "It's not enough that we win, everybody
else must lose," we believe that database users and Digital both win
when we accept and compete in a heterogeneous market.
That is why Digital's vision and strategy will continue to invest in
providing the world's best DBMS performance, integration, and
interoperability with Rdb and the AXP systems.
|