Event
Summary
On September
7th, 1999, International Business Machines Corp. (IBM) announced the Netfinity
4000R, a 1.75"- high server aimed at Internet Service Providers (ISPs) and Application
Service Providers (ASPs). The 4000R, which will have up to two CPUs (Pentium
II or III) per system, is a rack-mountable, stackable system. The system, expected
to ship in late September, will cost $3,000 to $4,000.
Market
Impact
This announcement
sets a new level of CPU density at "two per U" or 84 per rack (42U-high) - 75%
more than the previous density mark of 8 CPUs in a 7U height, or 48 CPUs per
rack. In addition, this server, based on the Network Engines WebEngine product,
can be clustered up to 256 nodes. The result of this could be a RAIS - Redundant
Array of Inexpensive Servers. At a time when everyone is jumping on the "appliance
server" (a/k/a "thin server") bandwagon, this product has the potential of having
a tremendous impact for providers needing this level of CPU power, and for web
caching situations.
IBM's entry
into this space will have a negative effect on companies like Cobalt Networks
(maker of the CacheRAQ thin server line) - the 4000R is a similar product, but
it now has IBM's corporate stability and strength behind it. Dell and Compaq
may feel the heat, but their appliance servers are considerably larger, with
more features, and play to a slightly different market - one where replacing
hot-swappable components, rather than replacing the whole server, is important.
This product
will provide growth for both IBM and the general market, by accelerating the
recent growth of thin servers. In keeping with the general server market trend,
it will lead to market consolidation as the small-manufacturer offerings get
overshadowed over the longer term (24-36 months, 60% probability).
User
Recommendations
This announcement
is a departure from IBM's heretofore "me-too" offerings, and may signal a shift
in IBM's focus from standard "vanilla" products to more leading-edge systems.
(Of course, if IBM had developed this on its own, rather than as a buyout, that
would be a definite signal of a change of attitude.)
This product
will allow a potential customer to pick a dense, thin server from a major hardware
vendor, thus reducing the risk of being left with an unsupported product. In
addition, IBM provides servers of all sizes and OSes, so a customer can have
an entire "Big Blue" computer center, rather than a thin server from Company
A, a mid-range Intel server from Company B, and a Unix server from Company C.
Finally, this
product raises the bar for CPU density, and we expect other companies to try
to match this. However, it will be another 6-12 months before we expect those
products would ship, giving IBM a big head start in the arena.