IBM Netfinity Server Division

  • Written By:
  • Published:

IBM Netfinity Server Division
R. Krause - May 15 , 2000

Vendor Genesis

International Business Machines was formed in 1924, although its origin lies in the formation of C-T-R, the Computing-Tabulating-Recording Co., in 1911. The company changed its name to IBM while under the leadership of Thomas Watson, Sr., who ran the company until 1956, when his son Thomas Watson, Jr. became CEO.

IBM's primary products in the 20s and 30s were, as its name suggests, business machines such as tabulators/calculators and typewriters. IBM's first "calculating computer" the Mark 1, was developed in 1944, but it was not until the 1950s and 1960s that IBM became more fully involved in computer production, eventually becoming the preeminent computer manufacturer in the world.

The System/360 (released in 1964) is widely considered to be the computer which established IBM's leadership and its dominant position. Since the System/360, IBM has gone on to be, at one time or another, the leader or a strong contender in most of the major computer market segments. These segments include mainframes, servers, desktops, notebooks, palmtops, storage systems, disk drives, printers (via Lexmark, formerly IBM's printer division), and software.

IBM was not the first to develop a Personal Computer, but the IBM PC (first shipped in 1981) established the PC industry. In addition, IBM's support (as well as some poor business decisions) helped Microsoft become the dominant software company in the world - some say the dominant force in the computer industry. In the meantime, the IBM PC has steadily lost market share to rivals such as Compaq and Dell, to the point where IBM has considered whether to remain in the PC market.

The PC Server, a subset of the Intel-based PC, has developed as a separate and distinct market from the PC. The server market has been dominated by Compaq Computer Corporation for almost ten years, although Dell Computer Corp. has recently been catching up to Compaq in terms of market share. During this time, IBM has not been a serious contender for market share leadership, but has consistently occupied a solid second or third position. The Netfinity server product line is IBM's version of the Intel-based server.

IBM as a whole derives its revenues from a variety of sources, but the key areas are hardware (43%), services (35%), and software (15%).

Vendor Strategy and Trajectory:

Traditionally, IBM's strategy has been to be a one-stop shop for all a customer's computing needs. In addition to a multitude of computing products of all flavors, IBM has earned a reputation for having a superior customer service organization. In recent years, they have re-tooled their strategy to focus on being the leader in providing e-business solutions for customers.

Although this note is focused on the Netfinity product line, which is Intel-based, IBM is definitely not a niche player. IBM has arguably the broadest offering of computer products in the industry, catering to almost every conceivable computing need. In the general-purpose server space alone, there are five distinct product sets: Netfinity, AS/400, RS/6000, S/390, and NUMA-Q. However, breadth of offerings does not necessarily translate into market leadership: IBM currently holds third place in Intel server market share, behind Compaq and Dell in both the U.S. and worldwide markets. IBM had occupied the #2 position in the worldwide market until recently, so this is a significant step back for them.

IBM's strategy is in flux. For the past 24 months its focus has been on producing servers that take the top scores in any/all relevant benchmark testing. For the most part, IBM has been successful. However, this benchmark/performance focus has caused them to lose sight of the goal in two key areas: price/performance and packaging density. Netfinity price/performance (as measured in terms like $/tpmC) is significantly higher than Dell and Compaq, its two main competitors. Note that higher, in this context, is a negative.

IBM has also gained a reputation for providing servers in larger packages than the competition. This is significant because the high end of the server market is moving toward a majority of systems being rack-mounted. Since equipment racks have a finite amount of vertical space, density of CPUs and general functionality is important. Netfinity servers in general are larger than similar offerings from competitors, leading to reduced density of CPUs and functionality in a rack.

In years past, IBM was the 800-pound gorilla in the computer world. In recent years, they have suffered some stunning defeats, including the effective collapse of the Aptiva (home PC) product line. IBM is still strong in a number of areas, such as the proprietary AS/400, storage, software (the #2 software company in the world), and mainframe computers. Although few if any companies are #1 at everything they do, the division across IBM is troubling.

It appears that much of its current success is drawn from the mainframe and legacy markets, with the weakness appearing in the more dynamic markets such as Intel servers and PCs. CEO Louis Gerstner took a severely weakened IBM and restored it to health; now his team needs to strengthen the weaker areas. Trying to emulate Dell Computer's dynamism might be a good thing.


Vendor Strengths


As mentioned earlier, IBM dedicated 1998-1999 to getting the top scores in every relevant benchmark test. It succeeded in that regard, racking up top scores in such diverse tests as WebBench, TPC-C, NetBench, SPECweb, MMB, and NotesBench. This is an enviable record, but it comes at a price: Netfinity servers will often cost significantly more than the competition for marginally superior scores. (See "Price/Performance" in the Challenges section for details).


The Netfinity group has been the beneficiary of research and development (R&D) work done for the AS/400, S/390, and RS/6000 groups - all products focused on higher-end markets than the traditional Intel-based market. This R&D work has migrated higher-end features down to the "lower-end" products. In addition, Netfinity products were the first Intel servers to have PCI "hot-add" capability under a Windows NT environment - at a time when most of the competition could not even provide "hot-swap" for PCI.

IBM has also focused on providing diagnostic capability so that the user can isolate numerous problems quickly.

Finally, IBM is the US patent leader, highlighting its strong commitment to basic R&D.

Service and Support

IBM has long been known for its superior service and support. This well-deserved reputation was enhanced when Dell Computer selected IBM in 1999 as their primary customer service provider. A large part of IBM's revenues are derived from services, we expect this to continue. IBM is currently on track to generate approximately $33B in services revenue. The services figure alone would put IBM in the top 25 of the "Fortune 500".

Although "size doesn't matter", it certainly helps in this situation. IBM can use this strength to great advantage, and does. Compaq Computer bought Digital Equipment Corporation primarily to get its services business, but we believe Compaq's services are a distant second to IBM's.

Vendor Challenges


IBM's superior benchmark numbers came with a hefty price tag attached. For example, when comparing TPC-C benchmark scores, IBM's $/tpmC often far exceeds offerings from Dell and Compaq. This translates to a Netfinity 7000 having a near-$200K higher price tag (~$595K vs. ~$405K) for a system which performed only marginally better than a Dell PowerEdge 6350 in transaction processing. This is a clear disadvantage when selling into a price-sensitive market.

Recently, IBM released tpmC results for an eight-CPU 8500R which bring it closer ($18.98 per tpmC) but it needs to close the gap between it and Dell/Compaq. [Note: results mentioned are still under review by the TPCC.]


Related to price/performance is IBM's price structure. Although IBM has recently started providing lower-priced offerings such as the Netfinity 3000, some of its pricing is still too high. An example ties, once again, back to the TPC-C benchmark testing. IBM's price for the 8500R, with a full complement of memory and processors, is almost $86,000. This is almost $25,000 more than a comparably equipped Power Edge 8450 from Dell. That's a lot of lettuce.

System packaging and density

IBM's competition, notably Dell and Compaq, have been striving for years to pack more and more electronics into smaller and smaller packages, IBM seems to be trying to buck the trend. A typical example is the Netfinity 8500R, IBM's eight-CPU offering. Whereas Compaq, Dell, and HP are all able to fit everything into a seven-rack-unit (a/k/a 7U, with 1U = 1.75") high enclosure, IBM chose to package their system in an 8U-high chassis.

One rack unit isn't much - until you realize that Compaq and Dell can now provide up to 48 CPUs in a standard 42U-high rack, and IBM can only provide 40 CPUs in the same space. This reduces the effective processing "density" by 17%, unwelcome at a time when floor space costs are increasing. IBM has started to address this through its relationship with Network Engines, Inc., designer of the high-density Netfinity 4000R. However, IBM needs to do more.


Vendor Predictions

IBM has indicated that it will now shift its focus from "performance at any price" to "price/performance competitive". We think this is a good thing, although it will need to work hard to overcome the prevalent perception of big, expensive systems.

There are indications that IBM has realized that the system packaging needs to get more in line with the competition. One of these indications is the Netfinity 4000R, which provides up to 2 CPUs in a 1U-high package - a stunning density increase, which has now raised the bar for the rest of the "Big Four". Compaq is attempting to respond, but their "Photon" server (which will match the 4000R's density) is not due out until the second quarter of 2000.

Vendor Recommendations

Reduce enclosure size

IBM needs to re-tool some of their products to get in line with the competition. Specifically, they can't continue to provide larger-than-"standard" enclosures when Dell and Compaq keep pushing dense product offerings. We believe that IBM needs to have at least 50% of their product line re-vamped for density by the end of 2000, and we would prefer to see more than 75%, including the 5500, 7000, and 8500R. (Naturally, the 4000R can stay.)

Be more price/performance competitive

IBM needs to review its entire supply chain, from design through product shipment, to see where costs can be reduced. By adopting (or at least attempting to adopt) Dell's direct sales model and all its nuances, IBM can expect to see significant reduction in manufacturing and distribution costs. A by-product of this model is a reduction in inventory and carrying costs, at least partly due to lower inventory cycle time. (Dell is currently turning over inventory more than 50 times a year.)

In addition to the manufacturing expenses, design engineering needs to look at cost reduction by way of low-cost (but not "cheap") design. This may mean taking a low-cost design and making it more robust, rather than taking a mainframe-like design and trying to reduce costs, as seems to be the current case.

User Recommendations

One-stop shopping, high reliability

As mentioned earlier, IBM has the most diverse set of product offerings available. Although this offers little benefit to Intel-only or Windows NT-only customers, it does provide a strong message for users who realize they need much more than a Windows-based infrastructure. With the added benefit of strong reliability and a strong service and support group, users who have a heterogeneous, multi-OS computing environment must consider IBM seriously. Low prices are a powerful incentive for customers to consider Dell, but Dell is not yet as proficient as IBM in multi-OS environments.

One caveat: customers should assess their OS/environment and application needs before deciding they need to go with a one-stop shop like IBM. Using a vendor who can provide these varied systems is the effect, not the cause.

Wait a little longer for re-tooled systems

As discussed earlier, we believe IBM will re-tool all or part of the Netfinity product line before 2000 is over, perhaps as soon as the first half. Users may want to explore this possibility with IBM, to see if any revamping is coming. If IBM is willing to discuss it, customers should try to get as many details as possible, and then compare the proposed products to the current offerings from Dell and Compaq.

If the hoped-for revamping is completed (including shipment to paying customers) less than three months before a user needs to buy servers, they are best advised to go with another vendor. This is because of problems inherent in any new model. Customers wishing to select IBM in the "cautionary period" should ensure they get performance guarantees from IBM sales, otherwise they should go elsewhere.


comments powered by Disqus