As published in HPCWire
With Oracle’s acquisition of Sun for $5.6 billion (net), the company is acquiring, among other assets, an estimated 8.4 percent stake ($627 million)
in the traditional HPC server market. The question arises: is this an asset that Oracle will keep/nurture as it moves to integrate Sun into its
ongoing product and operations strategy? Here are some thoughts on whys and why not’s for the proposition.
Whys
Revenue is revenue — We estimate that HPC server sales accounted for about 4.7 percent of Sun’s CY2008 revenue; additional HPC revenues also accrue
to Sun from storage and service sales. Oracle is looking for the “acquired business [that] will contribute over $1.5 billion to Oracle’s non-GAAP
operating profit in the first year, increasing to over $2 billion in the second year.” Although much of this profit could be had from operations
efficiencies — Sun’s FY08 SG&A expenses were roughly $4 billion and R&D $2 billion — 5 percent of revenue is still a lot of change to
make up.
New technology entry point — HPC is a traditional creator and test bed for new IT concepts and technologies. Sun’s current position in this market
would provide an additional innovation pipeline into Oracle.
Data driven HPC — Database and storage software accounted for about seven percent of the middleware mentioned in our 2008 site census. However we
believe that requirements for data-intensive applications are among the fastest growing areas within traditional HPC arena. Such applications range
from genomic searching to national security analysis. Thus there are areas of HPC that fit well with Oracle’s basic business.
Edge HPC — Use of HPC technologies for non-traditional applications would appear to be a good fit for Sun and Oracle technologies. These applications
tend to be both data and scale driven and have the availability and reliability requirements associated with commercial computing.
Why-not’s
Strategic fit — It appears that Oracle’s major reasons for purchasing Sun are: first, to acquire Sun’s software assets, particularly Java; second,
to be able to provide an end-to-end solution to enterprise customers running database applications; and third, to protect Solaris, which is the
top operating system for Oracle. These are all excellent reasons, but none has anything to do with traditional HPC, which could be viewed as more
of a distraction than a benefit.
Margins are margins — Although Sun’s HPC revenue is worth a look by Oracle, the company may find that margins associated with this business make the
opportunity much less attractive. In general (and without specific knowledge of Sun’s results) HPC margins tend to suffer from discounting to academic
and government customers, strict and expensive requirements for many high-end sales, high levels of competition, and low levels of customer loyalty.
Been there, done that — Oracle does have experience in the HPC market dating back from the 1980s when Larry Ellison and/or Oracle became major investors
in nCube — a massively parallel computing company with a hypercube architecture based system. The company moved into the media server business,
but did not prove a good match for Oracle Even after a decade or so, Oracle may not want to venture into the HPC world again.
Overall, (in my role as Dr. Doom) I think the why-not’s outweigh the whys and that Sun’s traditional HPC business will fade away over time. I would
expect Oracle to honor all ongoing contracts and continue to service its products as long as customers wish to renew the contracts. It is also
likely that Oracle will continue to support its lighthouse sites, such as TACC and Tokyo Tech, for the R&D and corporate exposure benefits.
That said, the strategic fit and margin issue are too great to overcome. It is difficult to see how traditional HPC would contribute to Oracle’s
overall strategy.
I do see Oracle as a current and future player in the Edge HPC market. However, I am not sure that a traditional HPC presence is necessary for the
company to play in the Edge arena.