|Predicting the Next 3 Years
in the SPARC Server Market - (2007 to 2009)|
Editor:- December 8,
2006 - as I predicted in an article
2 years ago the
SPARC market in 2006 has shown strong signs of recovery and even
Looking ahead there are some technology and
market trends which can increase SPARC server revenue growth in 2007 but also
some competitive challenges which could slow down growth in 2008 and 2009. But
the slowdown is not inevitable, and Sun's SPARC server growth can be
sustained for the whole period if Sun reacts to the competitive situation
SPARC success and growth factors - (2007 to 2009)
computing - end-users have bought into the concept that in most applications -
the peak performance (which is related to CPU clock rates) is less important
than the sustainable average processor throughput. So although Sun's SPARC chips
have lagged behind Intel Architecture chips in peak performance - the SPARC
chips have increased their competitive advantage in throughput. This has been
achieved because Sun packs 2 to 3 times as many processor cores in its
multi-core CPUs than any of its competitors.
That design concept has
also delivered the benefits of providing more CPU power in a single
motherboard while consuming less electrical power. As Sun's main target market
is telcos whose datacenters have run out of physical space - the current and
next generation of SPARC servers offers the most attractive way to increase
datacenter processor power. As one telco told me recently - "there's no
more room on the carpet for more servers."
In the time frame 2007
to 2009 - there will be a massive growth in telco and ISP infrastructure to meet
the needs of the new internet video market. Some mainstream broadcasters have
already started to make selected content available on the web - and from those
small steps - it is likely that broadband video on demand (narrowcast
transmission) will replace satellite and cable broadcast in the next handful of
years. That's an application in which predictable volumes of high throughput
computing can manage the job - and the ultimate in CPU clock rates is not
required. Many current video on demand pilot solutions are not scalable -
because the storage is too slow. But this problem is fixable with new server
farm architecture - (see below).
SPARC's biggest risk factor -
(2007 to 2009) SSD-aware OS from Microsoft
In November Microsoft
added solid state disk
support to its Vista operating system. That was aimed at speeding up and
extending the battery life of notebook PCs. When this capability is ported to
its enterprise server OS - it will mean that I/O bound wasted CPU cycles
(which currently waste 50% to 75% of datacenter capability) will be unlocked by
the addition of relatively low cost SSD accelerators. The effect will to double
application performance of server farms running the new OS - thereby canceling
out the advantages delivered by Sun's higher density of SPARC chips.
I expect work on that suspected Microsoft project to reach the public domain
soon - Sun will get a breathing space of about a year - before this SSD-aware
enterprise OS component starts to impact Sun's SPARC server sales - in the
first part of 2008. That's because the launch release only supports a single
SSD with upto 4G capacity - whereas most SSDs in use as enterprise accelerators
are in the range from 32G to over 1T. Another weakness of Vista's SSD support
at this time is that the host needs the RAM to be as large as the SSD storage
segment. That's a serious weakness for enterprise use - but no doubt it will be
tweaked in later versions.
Several years ago in a popular article
Why Sun Should
Acquire a Solid State Disk Company I warned Sun that it should start its own
SSD-aware program for Solaris - but I also admitted that server companies like
Sun, HP and IBM had a vested interest in keeping quiet about this technology as
it could reduce their server sales - which were flat or declining at that time.
I expected they would all wait to see who jumped first. Well - it was
Microsoft. So it's time for Sun and Linux to play catch-up.
how Sun reacts to this the SSD threat - it can sustain (or lose) its
competitive advantage in 2009. SSDs will double the speed of Sun's SPARC
servers too - and that's already being done by SSD magicians on a
customer by customer
Looking 3 years ahead is as far as I can go in this SPARC
market prediction. And the prospects look better than at anytime in the past 6
years. Just over 10 years ago I published an article called
Previewing the 2nd
Decade of SPARC Systems in which I invited SPARC industry leaders to make
their predictions for what would happen in the period 1997 to 2006 - now past.
Most were wrong. Some were right.
As we enter the 3rd decade of SPARC
in January 2007 - we have a lot of good things to look forward to.
Charting the Rise of the Solid State Disk Market
OSDL Lightens the Load
Editor:- December 4,
2006 - Linux developer OSDL has lightened its headcount by a third
according to an article in The Register.
What's bad news
for Linux is usually good news for Sun. OSDL
|SPARC history - 1987
is a Nashua, NH based extreme |
mobile computing solutions provider, capable
of filling nearly every requirement, including a
full line of mobile
UltraSPARC Solaris note-
books and mobile servers; as well as open
based Unix blade servers.
VARs - USA / Storage