| The Need
is Strong, the Upside Potential is Huge |
Sun Microsystems'
SPARC processors have been losing their performance advantages over Intel
Architecture processors since 2001. Sun's recently lanched servers based on the
new SPARC IV chips have left most analysts unimpressed.
Yes - it has
two CPUs in each chip. So what?
Intel and
AMD will be doing the same
soon, and their clock rates are much higher. Sun's rapid operating system tweaks
in Solaris mean that users can get a X2 speedup in many applications
immediately, whereas rivals HP,
Dell and
IBM may have to wait
another year for Microsoft
to support similar features in upcoming Intel Architecture chips. So Sun has a
short breathing space, in which it can live off
stories
of good TPC benchmarks. But the emperor has no clothes. Everyone knows Sun
is not as good as designing fast chips which include tens of millions of
transistors as Intel and AMD.
Sun doesn't have its own wafer fab, so
it can't attract the best chip talent to integrate new ideas with manufacturable
technology. The short term advantages of the "simple" RISC chips have
long since been eroded as CPUs get fat with caches and multiprocessing support.
But unlike Intel and AMD, who have to rely on 3rd parties to support new chips,
Sun has a secret weapon, its Solaris OS. Computer architecture has moved on from
the 1990s when buying the fastest clocked processor, and stuffing lots of them
in a box guaranteed a faster server.
If Sun can deliver
factory configured Solaris servers which include SPARC or Intel Architecture
processors and run popular applications like
Oracle 3 to
4 times faster than the biggest 4GHz servers from HP, then Sun's customers
will take note. Such products could not only stem the flow of customers away
from Sun to
Linux, but actually
reverse the tide. Sun's competitors will react, but their inability to tweak the
OS as cleverly as Sun has always done with Solaris, means that Sun could get a
12 month window when its products are the sizzlers which everyone wants to own.
Here's the trick. It doesn't need faster SPARC processors. Sun just needs to
implement solid state disk support into Solaris, and ship boxes which include
SSD as default options.
I believe that a $100 million strategic
investment by Sun in SSD technology could return increased revenues of over $1
billion within the first year. Oh yes. And the new products will be more
profitable too... Here are a few ideas to support this.
- Solid state disks can speed up server performance by 2 to 4 times.
There's plenty of evidence for this, see
solid state disks case
studies and articles for examples. But the SSD market is currently smaller
than it should be because operating systems do not automatically recognise SSD
devices and optimize them. Instead SSD vendors have been tweaking applications
on a case by case, customer by customer basis. That's slow and too people
intensive. It's like backing up your disk one file at a time from a keyboard
without an automatic script.
- Sun has a track record of buying storage companies to fix
weaknesses in its own internally developed product families. Previous storage
companies acquired by Sun include:- Cobalt Networks, HighGround Systems, LSC and
Pirus Networks. While this is not in the same league as
EMC, which has acquired 9
storage companies since Jan 2000, it shows a willingness to look outside for
strategic solutions. A solid state disk company acquisition could cost a lot
less money than most of Sun's earlier acquisitions. The upside potential is
high. See also:- Acquired
storage companies
- Sun has a track record of advancing computer architecture.
Computer architecture changed in
1992 when
Multiprocessing moved out of the scientific supercomputer market and onto the
desktop with Sun's SPARCstation 10. It was 5 years before the IA market caught
up with that. Then Sun did the same trick again when it brought 64 bit
processing to the masses in 1996. Today in 2004 faster clocked processors do
not guarantee faster servers. The other parts of the computing network have not
kept pace. SSDs
are a 20 year old technology which have been used by the
military to
boost system performance. It's time for that technology to come in from the
engineering lab and enter the mainstream.
Who will Sun acquire?
I don't have any inside knowledge of that. There are plenty of
solid state disk manufacturers
to choose from, and I know that a lot more are making their plans to enter this
market. We'll just have to wait and see. The market will be big enough for all.
...2 Years Later:- - I was wrong
about Sun being first to pick up the SSD torch.
In fact
Microsoft became the
first operating system vendor to incorporate SSD awareness in its Vista OS in
2006. However, Microsoft's implementation of flash SSD support was so bad - it
nearly killed off the fledgling
hybrid disk
market. | |
... |
| Nibble:
Operating System Support Can Work Wonders |
Embedding native solid
state disk support into Sun's Solaris servers would have far reaching
effects:-
- Sun's SPARC servers would reach a
performance level which would take them 2 years ahead of the current SPARC CPU
roadmap, and at much lower development cost.
- Sun's AMD Solaris servers would run
typically 2 to 3 times faster than competitors' Linux servers using the same
CPUs.
| |
| . |
 |
SSD Bookmarks
suggested by - Woody Hutsell, President Texas Memory Systems |
Here's an article written by or
about Texas Memory Systems
Flash SSD Reliability (pdf)
Woody
Hutsell says he chose this article because Flash reliability is a topic of great
interest right now, and this paper approaches the subject in a unique and very
readable manner, starting at the chip level and working up through the board
level all the way to the enterprise architecture perspective.
Other SSD article suggestions...
Woody Hutsell says - "As
you know, the Storage Networking Industry Association (SNIA) started a new group
last summer, called the Solid State Storage Initiative (SSSI). This new
organization has the ambitious goal of becoming a major advocacy group for solid
state storage. SSSI is starting to publish
research
papers on solid state storage topics. The first one,
Solid State
Storage 101 (pdf), is interesting for the fact that it is the outcome of a
collaboration between many companies who, in most other settings, would be
serious competitors."
Other SSD bookmark suggestions...
StatspackAnalyzer - is a website
where IT professionals can paste their Oracle statspacks or AWR reports and get
analyses and recommendations for storage performance improvements. (It's free to
use but registration is required.)
Woody Hutsell says he
recommends this bookmark because - "A key ingredient to greater SSD
adoption is a better understanding within the user communities of just how
important storage performance is to mission success. StatspackAnalyzer.com isn't
a large website, but it does have some information, a forum, and even the entire
Analyzer rules list available for comment and improvement."
Editor:- thanks Woody for sharing your SSD links.
see also:-
Texas
Memory Systems - editor mentions on StorageSearch.com | | |
| . |
|
 |
Can You
Trust Your Flash SSD's Specs & Benchmarks? |
No - sadly you
can't! There are many intrinsic technical reasons why you
can't believe most published benchmarks for flash SSDs
(whether done by magazines or vendors) and why even the
tests you carefully do yourself don't give reliable
results which correlate with how the SSD will perform in
real-life applications.
We warned you of it this
problem here on StorageSearch.com last year - and now other
publications and vendors are starting to take it seriously
too. ...read
the article | |
| . |
| . |
| |