home
***
CD-ROM
|
disk
|
FTP
|
other
***
search
/
Power Tools 1993 November - Disc 2
/
Power Tools Plus (Disc 2 of 2)(November 1993)(HP).iso
/
hotlines
/
cpethl
/
sunqa
/
sunqa.txt
< prev
Wrap
Text File
|
1993-09-17
|
14KB
|
259 lines
H P C O N F I D E N T I A L
=============================
Written by Todd Thiemann
General Systems Division, Cupertino
SUN ARGUMENTS & TYPICAL QUESTIONS
=================================
Sun is presently making an aggressive attack with the
SPARCserver 1000 (SS1000) and SPARCcenter 2000 (SC2000), but
most of what they're touting is futures and marketing hype. What
follows is Sun's typical sales pitch (A for Argument) followed
by a counter argument (CA) that points out the weakness in Sun's
position and the strength of HP's offering. Following the
counter-arguments are questions for Sun to answer along with Sun
weaknesses to emphasize.
A: Sun's SMP technology provides superior fault resilience to
HP's Series 800 by permitting the system to reconfigure around a
around a failed processor.
CA: Sun's MP resilience buys you nothing because more CPUs with
the same performance means increased points of failure. HP's
PA-RISC Uniprocessors have superior performance & fewer
components. HP's 890 Corporate Business Servers can also
reconfigure around failed processors, but we have seldom
encountered such problems (HP's average MTBF is 2.5 years!).
MP fault resilience is Sun's desperate attempt to resuscitate a
dying processor, and this is functionality HP has had since
1991.
A: SS1000 and SC2000 SMP performance is better than HP's
Series 800. Sun is the performance leader.
CA: Sun's best Uniprocessor benchmark is 108 TPC-A using a 40
MHz SPARC CPU in a one-way system. The 50 MHz chip in
an 8-way system produces 408 TPC-A. Does this demonstrate
good scalability? Is Solaris the limiting factor? What will Sun
offer and when? Why isn't it available today? HP's Model 70
Nova's provide 411 TPC-A with only 2 CPUs (better than Sun's
8-way!) and Emeralds produce an audited 711 TPC-A today
and about 1600 TPS in the next 890 release. Sun's SPARC
CPU is at the end of its life, and Sun is trying unsuccessfully
to extend that life through MP technology.
A: Sun provides easier & cheaper upgrades; future upgrades to
faster CPU chips are simple.
CA: Sun locks you in to a inflexible upgrade path dictated by
Sun Microsystems. To add more I/O slots, you must add CPUs &
memory. HP offers upgrades with greater flexibility within
machine classes (F, G, H, I) and between machine classes. HP
also allows you to build on your existing hardware investment.
Did you know that you cannot use most Sun disks or memory
on the SS1000? Or that SS1000 CPUs and internal disks
cannot be used with the SC2000? With Sun, you lose your
investment as you expand.
Look at current offerings. Can Sun provide a SC2000 with 20
processors TODAY? No. And while the SC2000 maxes out at
138 GB of disk (moving to 1 TB in April 1994), 890 Corporate
Business Servers can manage up to 1.3 TB of disk NOW.
A: How much of HP's installed base uses client-server as opposed
to HP's traditional emphasis of terminal based minicomputers?
Does HP really believe in client-server computing or is HP
clinging to its 20 year-old timesharing model?
CA: Sun is talking through both sides of their mouth - while
spouting "client-server", they have quietly sold terminal
controllers. HP listens to customers and helps solve their
problems. We support the gamut from client-server
architectures to ASCII terminals to offer customers the biggest
bang for their buck. Sun continues to impose its computing
vision on customers. Why did Sun wait until 1993 to offer
X-terminals?
A: The SPARCcenter 2000 provides a superior enterprise server
compared to the HP 9000/890. It provides 5 times more CPU
expansion, 1000 Gb of storage and 2.5 times more RAM (up to
5 Gb).
CA: Always compare current Sun capabilities and current HP
capabilities. Or futures to futures. In either case, HP wins.
Look at the current situation:
Aug93 HP9000/890 SC2000
===== ========= ======
CPUs 4 8
TPS 750+ ~500
Max Disk 1300 GB 138 GB
Max Memory 2 GB 2 GB
Max I/O 112 slots 4-10 slots
A: With AT&T/NCR, Sun, and Novell running the advanced
SVR4, does HP plan to offer SVR4 functionality? Solaris
offers SVR4, and SVR4 means more functionality.
CA: HP-UX is a robust and proven OS that will continue to evolve
to meet customer needs without pain such as Sun imposed on
its installed base in moving from Solaris 1.x to Solaris 2.x.
HP will continue to improve on HP-UX by adding appropriate
technologies from sources such as OSF, USL, and other
technology providers.
Many customers tell HP that the entire computing environment
is more relevant than simply the OS. HP provides the most
robust environment with functionality unavailable on Sun
through products such as SAM, Switchover/UX, OpenView, &
Operations Center.
A: It is likely that OSF and UNIX International will merge
technologies. If this is true, HP will have to migrate its
customers to another OS. When? At what pain? Or does HP
plan to offer its customers old O/S technology with HP-UX
based on SVR3?
CA: OSF and USL are converging on a common set of definitions
(SVR4 is AES compliant, OSF/1 is SVID3 compliant). We
combine the best of both in a smooth evolutionary path. HP has
not and will not impose painful migrations on its users such as
Sun forced in moving from Solaris 1.x to 2.x, a move which
requires extensive application rewriting.
A: Sun technology is demonstrated through its outstanding I/O
Pipeline with XDBus and Dual XDBus - fast systems with room
for growth.
CA: Overall system performance measures such as TPC and
LADDIS are better yardsticks than single dimensions such as
bus speed. Sun's TPC numbers significantly lag HP's. HP's 890
bus is four times faster than Sun's, and Dual XDBus is a long
way from being deliverable. HP has offered packet-switched
technology since 1990, something Sun introduced in 1993!
A: Sun has close RDBMS vendor relationships (Oracle) - closer
than HP.
CA: HP is now Oracle's largest hardware platform and has been
the UNIX sales leader for the last five years.
QUESTIONS FOR SUN TO ANSWER
===========================
Will you be able to upgrade your Sun systems as your computing
needs grow? Can you upgrade while maximizing your investment.
Point: HP permits customers to move within the F, G, H, and I
series while maintaining their disk, I/O card, tape drive and
memory investment. The comparable move within Sun's product
line (SS10 ===> SS1000, 690MP ==> SS1000, SPARCserver
Classic ==> SS1000) abandons the customers investment in
disk and memory. Also, customers moving from a SS1000 to
the SC2000 lose their internal disk investment.
Can Sun continue to improve SPARC performance as HP has improved
PA-RISC performance?
Point: Most industry analysts believe SPARC is at the end of its
lifecycle. HP improved PA-RISC performance at a rate of 60%
per year, and has vowed to continue improving at 70% per year
while Sun is struggling to keep SPARC ahead of Intel's
Pentium. For instance, their latest SuperSPARC chip was 18
months behind schedule.
What assurances do I have that Sun will not liquidate my
computing investment in the future?
Point: Moving from a SS10 => SS1000 requires new memory &
internal disks and SS1000 => SC2000 requires new internal
disk. An analogous HP migration (within the F,G,H, & I series)
does not eliminate your disk investment. As further evidence
of Sun's lack of regard for its installed base, Sun caused
tremendous trauma to users who will have to rewrite
applications when migrating from Solaris 1.x to Solaris 2.x.
HP strives for backwards compatibility with innovation.
Who offers reliable service, support, and consulting after
purchasing your hardware platform?
Point: HP can support you in the event that something goes awry. Sun
has not devoted significant resources to post-sales support.
Sun has outsourced a much of its support to third parties.
HP's field manpower is at least 10x Sun's manpower.
Can Sun provide 10 reference accounts where the customers are
running strategic business applications on SMP servers? Not
file servers but key application servers.
Point: Sun does not have HP's experience gained from 21 years
offering commercial systems (13 years with UNIX systems, 4
years with SMP technology). Sun does not offer the robust
commercial environment that HP provides (CA-Unicenter,
OpenView, etc.). HP reference accounts include British
Telecom, Cheeseborough Ponds, Dunn & Bradstreet, GTE,
Hammersley Iron, Hughes Space & Communications, Fritz
Companies, Sony, Texaco, Timex Corp (for additional
information, refer to the MFA HOTLINE or to the PowerTools
"Customer Reference Materials" folder).
What are reliable performance numbers (TPC-A & TPC-C) on
available SPARCserver and SPARCcenter systems (not beta OS code,
but PRODUCTION systems)? What are the systems' overall
reliability figures (MTBF)? What does Sun's SMP "fault
resilience" do for their MTBF?
Point: Sun cannot deliver robust, reliable solutions TODAY. Solaris
2.3 will not be available in production versions until October
or November of 1993.
Why does Sun still push ONC+ when the industry is standardizing
on DCE? Where is Sun's real commitment to industry standards,
particularly when the standards chosen or developed are not
based on Sun technology?
Point: Sun does not adhere to industry standards and prefers to lock
customers into Sun's particular implementation. Sun products =
Sun proprietary.
Sun touts its fault resilience, but what is Sun's capability in
the areas of:
o Automatic power fail recovery
o Remote diagnostics
o Performance tuning & capacity planning tools
o Commercial-grade print spooling
o Distributed system backup over the network
Where are reliable measures of Sun's production system
performance?
Point: Sun often quotes SPECrate_int92-type hardware benchmarks-
artificial measures for hardware only that do not measure
commercial performance. A reliable way to estimate machine
performance is to run industry accepted benchmarks like the
TPC-A and TPC-C benchmarks. Sun avoids performing such
benchmarks on its SMP machines and instead relies on esoteric
hardware benchmarks that do not indicate real-world
performance. When Sun provides benchmarks, they often use
pre-release software such as Solaris 2.3.
What is innovative about Sun's XDBus?
Point: Sun touts its XDBus as providing superior I/O performance to
"circuit-switched technology currently used by most vendors."
Sun introduced "packet switched bus" in 1993 as offering
superior I/O performance - HP has used packet switched bus
technology since 1990. Sun's XDBus has a peak speed of 640
MB/sec, and a sustained rate of 500 MB/sec while HP's HP-PB
bus offers 1 GB/sec.
How can you effectively manage a multivendor datacenter using
Sun systems?
Point: HP has solutions available today to manage heterogeneous
datacenters (OpenView, GlancePlus, OmniBack). Sun has very
limited offerings.
How many ISV/VAR applications are available on Solaris 2.x?
Point: Only 600-850 applications will be available by the end of 1993 -
not even close to the thousands that are available on HP-UX
9.x. ISV's don't want to make the port to Solaris - they want
to stay with Solaris 1.x. Proof of this can be seen in Sun's
releasing SunOS/Solaris 1.1C for the SPARCclassic and LX
systems that were supposed to only support Solaris 2.x and
never support Solaris 1.x. ISV's see little benefit in Solaris
2.x and are sticking to Solaris 1.x.
HP-UX 9.x offers demonstrated maturity & reliability while
Solaris 2.x is still brand new and has yet to shake out bugs
inherent in any new operating system. Do customers really
want to be beta sites working out Solaris bugs?