31 , 2000
The Intel-based Server (also known as "PC Server") marketplace is encompassing
an increasingly broad range of uses, functionality, and performance. Within
this large market there are a number of smaller markets or segments based
on such things as applications (e.g., Web serving or messaging) and functionality
of the server itself (general purpose server vs. server appliance).
most common way of segmenting the market is by size of user population
(Workgroup-Department-Enterprise-Large Enterprise), which can be loosely
translated into the amount of processing power, which usually equates
to number of CPUs. In this note we will examine the lower end of the spectrum,
that of Workgroup/Department servers. Although we could expand the definition,
we will limit discussion to servers having no more than two CPUs.
This market serves a number of needs: small companies that need a workhorse
server, larger groups that need a system to interface to a larger storage
array, departments that need an application server, enterprises that need
a file or print server. In all cases, more functionality is needed than
a typical server appliance (i.e., a focused-functionality server) can
provide. The target buyer is the manager of a small office-home office
(usually called SOHO), or the operations manager of a large department
in a large company, or the CIO of a 25-50 person startup. In other words,
anyone buying for a group of less than 100 or so users.
In this market segment, average prices vary from $1500 for an entry-level
server, to more than $15,000 for a fully-loaded departmental server unit.
Because the prices are significantly lower than those for enterprise servers,
sales cycles tend to be less - weeks instead of months. Note that as users
add more equipment such as disk arrays, the price for a full system can
exceed $100,000. The market has traditionally not considered a server
to be in the "mid-range" until the cost reaches the $100K-$1M price band.
However, that classification applies to entire server market, including
very-high-end machines such as the IBM S/390 and the Sun "Starfire"/E10000.
In the Intel market segment, the "mid-range" segment starts around $7000
for a machine without a lot of add-ins.
present market size is approximately $20 Billion worldwide, and grew about
10% since this time in 1999 - slower than the general server market, but
not bad overall. In this market, there are two clear leaders Compaq Computer
Corporation and Dell Computer Corporation, two challengers Hewlett-Packard
and IBM, and a bunch of also-rans who are presently fighting for a piece
of the 25% of the market that the "Big Four" don't yet own.
Depending on which metric one uses, this market segment is both expanding
and consolidating. The segment is consolidating in terms of the number
of vendors and alternatives available (within a vendor's offering). As
discussed above, the number of vendors in the general-purpose server market
is decreasing. Regarding the alternatives available: as recently as two
years ago, there were a number of uniprocessor servers available from
the "Big Four". At this writing, there are only three available (four
if you count Compaq's NeoServer, which we consider a special case). At
a time when the incremental investment required for a dual-processor-capable
model is so small (a few hundred dollars in some cases), the appeal of
a bounded-solution uniprocessor will be limited to home offices and the
like who are trying to save every dollar.
In addition, servers at this end of the price range are very low margin
units - less than 20% margin, as compared to a high-end machine's 40%-50+%.
The combination of these factors will lead to the eventual disappearance
of the general-purpose uniprocessor. (Please note that we do not include
server appliances in the preceding comment - they are special cases, and
do not yet follow the standard market trends.)
In an absolute sense, the market is expanding - annual growth is 10+%.
However, this compares to growth of 20+% for the server market as a whole.
The areas where we believe the small server market will have its greatest
success are the SOHO (Small Office-Home Office) market and Asia. Sales
patterns in Asia indicate the region is generally more price sensitive
than the U.S. This makes the small server a "natural" for the region.
In addition, growth in the Latin American (LatAm) market indicates a potential
good fit for small servers.
As server appliances and their applications proliferate, this segment
will come under increasing pressure. We do not expect the small server
market to disappear, but we expect there will be further consolidation
of vendors and products as more-targeted systems become easily available.
In the small server segment, as with the general-purpose segment as a
whole, the leaders are Compaq, Dell, HP, and IBM. In the general-purpose
market, these four vendors accounted for approximately 75% of the Worldwide
unit sales volume and approximately 80% of the U.S. volume in 1999. As
noted previously, the market consolidation trend that started more than
three years ago has continued. Among the "Big Four", Compaq and Dell have
separated themselves from the pack in the U.S.: their share percentage
has increased from approximately 52% in CY 1998 to approximately 60% in
CY 1999 [Source: IDC].
Compaq (ProLiant product line)
In general, Compaq's mix of I/O, hard drive, and system level features
is a good amalgam of features. Although Compaq may not have the "killer"
feature in any one area, we believe they are well-rounded and thus highly
suitable for customers seeking a general-purpose machine.
Although this is not a product strength per se, Compaq's leadership in
the market gives it a little more leeway, especially if Corporate is still
trying to figure out how to sell systems through direct sales. However,
if Dell closes the current 12% gap (in US unit sales market share), the
value of this strength decreases to near zero.
IBM is the clear leader in this category, Compaq is a strong #2. Compaq
absorbed Digital Equipment Corporation's extensive service group when
they merged in 1998. Although this group has lost a little "shine" in
the last two years, it is still strong.
In an oft-referenced survey done by Computerworld magazine in November,
1999, Compaq's reliability (as rated by IT professionals) was among the
lowest of any vendor. This kind of publicity is a bad thing for Compaq.
They need to determine if the survey results were due to real quality
issues, or just perception. We believe it is more than perception, and
would suggest that Compaq refocus in improving the quality of their systems.
Recent survey results from Technology Business Results (an industry benchmarking
firm) indicate that Compaq is improving in this area, but it takes a lot
to overcome negative perceptions.
As we have discussed before, Compaq is trying once again to develop their
direct sales model. This effort is one of the high priority projects within
Compaq, because they know it could realize cost savings approaching 15%.
Unfortunately for Compaq, they also knew that the last time they tried
it, and the time before that. We believe that this is their last chance
- they either get it right (or darn close), or they suffer the consequences.
In this case, the consequences are Dell passing them to take the unit
sales leadership position.
(PowerEdge product line)
Sales Cycle and Delivery
Dell continually refines their direct sales model, and they consistently
feedback customer data to the development organization. One of their latest
features: focused, packaged systems. These are preconfigured systems aimed
at particular segments, such as messaging or e-commerce, and Dell even
breaks it down by number of clients/licenses/seats a customer expects
to serve. Although we expect there to be a few early bugs vis--vis system
tuning and tweaking, we feel this methodology has tremendous potential.
In the aforementioned Computerworld survey, Dell scored very high for
reliability - not as high as IBM, but still quite good. Part of this comes
from the constant feedback generated from the direct sales model, which
is then fed back to help correct problems.
As stated often, both in TEC notes and elsewhere, the direct model allows
Dell to charge lower prices. Part of this is from cutting out the "middleman",
i.e., VARs/distributors. In addition, the high inventory turnover means
a lower cost burden on the product. The other big manufacturers have started
reducing prices, but Dell is still the one to beat.
Research and Development
As a percentage of revenue, Dell's R&D spending is about one-third that
of the rest of the Big Four. Dell does this by forming close relationships/alliances
with companies whose technology Dell can leverage. The most well-known
example is Intel (we believe it was Bill Joy of Sun who once referred
to Dell as Intel's distribution channel), but Dell is happy to work with
whomever can help them accomplish their goals.
Hewlett-Packard (NetServer product line):
HP was among the earliest adopters of redundant, hot-swappable components.
They maintain this policy through much of their product line, migrating
these features down to some of the smaller servers such as the NetServer
LC 3, LC 2000, and LH 3000. Although some of this functionality may not
be strictly necessary in this class of machine, it is nice to have it
available if desired.
HP has a history of quality packaging design, with an extra focus on ergonomics.
We appreciate this, and see evidence of its migration into the lower-end
machines. However, we also feel that HP sometimes underutilizes space.
An example is the LPr: a year ago, two hard drives in a 2U box was "ahead
of the curve"; now, Dell puts four hard drives in a 2U enclosure. Although
we believe HP needs to refresh the system, we still like its design.
Weakening Market Position
HP presently finds itself in the #3 U.S. position (just barely) and #4
worldwide for unit sales market share. Much of this decline is at the
hands of Dell. HP is by no means out of the race, but they need to make
a concerted effort to regain some share.
Half of their current product set (in the 1-2 CPU space) is at least two
years old. Although the server market is less "energetic" than the desktop
market, this is still a long time. Even the LPr is 18 months old - when
it first appeared, it had a groundbreaking feature set, but now it is
getting to be just one of the crowd.
Normally we wouldn't comment on this, but: HP's marketing people need
to be more attentive to the image the NetServer section of the website
presents. We don't know if it's the take-no-prisoners marketing hyperbole,
or if it's just the selective use of "facts" that concerns us. For example:
the performance page for the NetServer LPr - actually a system we like
- talks of having the best performance on the SPECweb96 benchmark,
using Microsoft's IIS as a web server. (The purpose of the testing was
to demonstrate a lower-cost alternative to Unix systems.) What HP neglects
to say is that when they tested, their results were the only ones from
a major vendor using IIS 4.0. [Note: IIS 4.0 improved performance over
IIS 3.0, which had been used by others of the "Big Four". Subsequent IIS
4.0 testing surpassed the LPr's figures.] We understand that all companies
try to put their products in the best possible light, but our preference
is that companies keep the well-tailored scenarios to themselves.
In addition, significant parts of the NetServer section are so out of
date that we wonder whether anyone in the company actually cares about
maintaining it. (They may, but the fact that it is so noticeable is significant.)
Combine that with year-old performance figures, and a bad case of the
lazies becomes apparent. In fairness to HP, we have noticed an improvement
recently, but we still see this as a challenge to be overcome.
Benchmark of the Standard Performance Evaluation Corporation
(Netfinity product line)
IBM has earned a reputation for high reliability, as evidenced by their
strong performance in a Computerworld survey in November, 1999. Although
the focused more on the higher-end machines, we feel this strength is
included in the entire product line. The reliability focus comes in part
from the migration downward of features and functionality from the mainframe
side of the business (whence came many of the Netfinity design team).
IBM has one of, if not the strongest customer service and support organizations
of any computer manufacturer. There is not much we can add to that.
In 1999, IBM set out to garner the top scores in every reasonable benchmark
they could find. They were successful for the most part. With these performance
results, IBM is now ready to move on to other areas, such as reducing
their price/performance figures (historically a problem for their enterprise
IBM has a whopping eight products in the uni- and dual-processor space.
The next greatest number is six, from HP, and we believe one or two of
those will be retired in the next 12-18 months. There is no way that this
market segment justifies the breadth of offering - why would anyone take
the time to differentiate between two machines whose range is so bounded?
Price and Price/Performance
Although IBM has taken steps to reduce price in the last year, we still
believe they are at a disadvantage, especially when compared to Dell.
However, we believe the pricing problem is more prevalent in the high-end
machines (four and eight CPUs), and we believe IBM will have "solved"
the problem within 12 months.
Systems Too Big
As is the case with IBM's enterprise servers, the rackmount enclosures
are just too big. Some of them can house more hard drives than the competition,
but the absolute tallest a rack unit should be is 7U (1U=1.75") - IBM
has three models at 8U. IBM is due for a product refresh in Q2CY99, we
hope/expect they remedy this problem.
The current market "losers" include Acer, Gateway/ALR, NEC/PB, Fujitsu/Siemens,
and Micron. Although they currently make some money selling servers, there
is a very high probability (>90%) that none of them will ever see 5%.
This has very little to do with technology, it is more a matter of marketing
and mindshare lost to the Big Four. Acer has recently declared it intends
to get out of the hardware business to concentrate on software, although
not immediately. Gateway bought ALR in the hope that it could translate
its desktop success into the server space, but it hasn't worked out. Micron
bought NetFrame ostensibly for the same reason, with the same result.
Fujitsu/Siemens has approximately zero market presence in the U.S., and
we believe NEC/Packard-Bell's server days in the U.S. are numbered. In
short, we expect little market impact from any of these vendors.
We expect this market to grow at a 20%-25% annual rate over the next three-five
years. This market is slightly less susceptible to the forces that have
given rise to such new segments as server appliances. Companies and workgroups
will continue to need general purpose computing; server appliances are
aimed more at the larger GP servers, where a customer is paying for computing
"overhead" which is not really needed. In smaller environments - departments,
workgroups, small businesses (less than 100 employees) - most/all the
functionality of a general purpose server are utilized. However, will
still expect to see pressure from appliances in certain targeted areas,
such as groups that only need a print server. We do not believe this segment
will be consumed from above (enterprise) or below (PCs and appliances),
but things might get sticky at times. We also do not believe this market
will consume/subsume other markets - vendors view these servers as necessary
evils, and will not try to shepherd customers from other segments into
also expect margins to get tighter, which will result in the remaining
contenders being forced to find other ways of making money on these boxes.
Such methods might include innovative design to reduce inherent cost of
manufacture, switching the manufacturing location "off shore" (as many
have already done), and bundled application solutions (a la Dell).
Once a company gains a reputation for lower quality/reliability - whether
deserved or not - it becomes difficult to change that image. Compaq has
not yet suffered significantly for this perception, but we have found
the market share effects tend to take a little longer to be felt, but
also take longer to dispel. Compaq should redouble its efforts in both
engineering and marketing (not sales) to ensure that they become known
as a reliability leader. Although this will take at least 12 months to
effect a turnaround, the benefits will be worth it.
addition, Compaq should redouble its efforts for direct sales, especially
over the Internet. Although direct sales is not the sole reason Dell is
catching up to Compaq, it is a big one. If Compaq can implement their
program well, it will help eliminate a big competitive selling point for
Dell should keep doing what they're doing, but try to leverage some more
technology. Every time Intel has had a hiccup (e.g., the "Coppermine"
chip or "Saber" main logic board problems) Dell has suffered a little
from the negative effects. Dell is still loyal to Intel, and AMD does
not yet have a viable offering for servers, but the ups and downs are
no doubt frustrating to Dell. Investing a little more in R&D may provide
a way to mitigate the effects of relying so heavily on Intel.
Continue the product set revamp started with the LPr and continued with
the LH 6000 and LT 6000. If design engineering can come through in a similar
fashion for the rest of the product set, HP will have a much stronger
competitive offering. We would suggest HP look at retiring the LH3 series.
(HP says it has been retired, but it still appears as a current product
offering on their website.)
Get whomever is in charge of the website to pay more attention to it on
a timely basis. The obvious inattention makes a high-class company like
HP look like a bunch of tyros.
should shrink the boxes, and work on price/performance. Recent products
indicate that IBM is starting to come around to the "smaller is better"
and price/performance mindset. IBM should be due for a product line refresh
in April: if the anticipated refresh addresses these concerns adequately,
IBM should see increased interest in its Netfinity products
Long Term Winners, Challengers, and Losers
believe Compaq and Dell will be the long-term winners, although not necessarily
in that order. Dell will win through pricing, delivery, and customer satisfaction.
Compaq will maintain its lead by improving its real/perceived reliability,
but mainly by matching Dell on the product delivery side of the equation.
Compaq has had enough chances to "get it right"; if they fail this time,
a good chunk of the market will write them off. Although it will be difficult,
we believe Compaq will eventually succeed. (70% probability).
believe IBM and HP will continue to be strong challengers, although they
need to take some of the steps described above to remain viable (in market
terms). We believe they will stay about even in terms of market share
(75% probability), unless one of them makes a serious blunder (<20% probability).
We believe that the "losers" mentioned earlier will not improve their
standing in the market, and will cut back their efforts (80% probability),
unless one of them develops an astounding new technology which they can
leverage into a new, highly-advanced product (<10% probability).
The type of user who cares about this market is one who needs modest functionality
at a good price, and a system that is geared toward small environments.
Users who are expecting to experience rapid growth may find it more suitable
to purchase a stripped-down, upgradeable quad-processor unit, or a dual-processor
that can be easily upgraded to a quad.
general, users should first decide which features are important to them,
compare the vendor offerings having those features, re-assess their needs
versus the products available, then cull the list to three (at most four)
specific products. Decision factors should include features, price, reliability,
long-term costs, and service/support.
As mentioned earlier, these products are suitable for small-to-medium
departments and SOHO, medium-to-large workgroups, small businesses. These
products are not for server consolidation, rather they are more for providing
basic functionality (network, file/print, storage, applications)
Below is a set of suggestions for users to investigate further, for different