by
Zsolt Kerekes,
editor - StorageSearch.com
This article - written and published just before 2009 began - offered my
speculations about the important trends we might expect to see in 2009. It was
updated throughout 2009 with snippets related to those predictions. For a
timeline and summary of what did happen in the SSD market in 2009? - see
2009 SSD history.
2009 - Year of SSD Market Confusion?
When I
summarized the most important
developments in 2007
- I called it the "Year of SSD Revolutions". Solid State Storage
remained the most exciting and fastest changing part of the
storage market in 2008
too. I wondered - is there a simple way of summing up the most important
SSD market
developments in 2008? And just as important - is it possible to make any
useful predictions for what may happen in 2009?
In many ways
the SSD market resembles the wild west - with a range of completely different
technologies, architectures, interfaces and form factors all clamoring for
attention - and saying "Look at me! I'm best."
So - in 2008
- it wasn't surprising to see some would-be sheriffs riding into town - in the
shape of various industry
trade ORGs - trying to bring some law and order into the chaotic new SSD
territories. These included the
SSD Alliance ,
JEDEC and
SNIA. They may achieve
some measure of success in the long term - but they're recently arrived
strangers in this town - just like everyone else.
I think the
lawman's lonely task in the wild west SSD town can be likened to that of the
wrangler he meets in the stable.
The wrangler has been told to
brand all the horses in the livery. Only trouble is - they have already
bolted. So he's got to round them up again. But he faces the problem that no
one can agree on what the boss's horses actually looked like - or which way
they were headed when they ran off.
I talk to a lot of
SSD companies, users,
analysts and
VCs looking at the SSD
market - and one thing that is clear - is that there is no single technology,
vendor, approach or solution that is going to dominate the market. I should
mention here that not everyone agrees with this. And in fact several vendors
have already announced they each expect to get 50% of the future market for
SSDs. In my view those claims are ridiculous and mostly go to show how little
the SSD product marketers in those companies actually understand the
complexities of this market.
Instead of a homogenous SSD market
dominated by BIG- BRAND-X - I think you're going to see what looks like a
confusing heterogeneous picture developing in which multiple SSD solutions
co-exist in the market - and apparently address identical application needs.
There will be a lot of like and unlike SSDs which can do the same job - but
which (internally) are architected in completely different ways. I explained
some of the reasons for this - in a recent article about the
rackmount SSD market.
And you're going to see similar confusion of choice all the way down to
standalone SSD chips
for mobile devices. As I've said in many recent discussions - customers have
widely different risk / reward / competence / convenience profiles - so there's
no such thing as a single ideal solution. That's what makes this market so
interesting.
Translating this into specifics - it does mean that you're going to
see proprietary and open architecture SSD arrays going for the same market. You
are also going to see computer bus SSDs (like
PCIe) and storage bus
SSDs (SATA,
FC etc) apparently
targeting similar customers. And you will see an even greater proliferation of
form factors - including more flash storage mounted directly on motherboards -
in notebooks (such as Intel's
Z-P230)
and servers too (such as flash
EcoRAM
from Spansion) . And
you're going to see more choices in complex OS-agnostic but value conscious
hybrids.
So
what else happened in 2008? - Although most of the SSD market developments in
the year were evolutionary rather than revolutionary - there were some real
surprises - even for me.
Back in 2007 the best ratio of sustainable
reads to write IOPS in fast enterprise flash SSDs was about 10 to 1. I
thought it would be another year or so before that got much better than 5 to
1. But I was wrong. Because in 2008
Violin Memory launched a
flash SSD which closed that gap down to 2 to 1 and also introduced the idea
that a flash SSD can do an immediate (non blocked) read from a virtual
address region which has just been written to. I thought 2 to 1 was a major
advance until Fusion-io
unveiled more details of its own SSD - which pretty much delivers R/W IOPS
parity.
Once you learn how these things have been achieved (often
under NDA) the overall product features make more sense. In several phone
conversations with vendors this year - I've had Eureka moments - and commented
on the cleverness of their design "tricks". I've rightly received mild
verbal chastisement - explaining that many of these "tricks" are
actually the subjects of patents - and that a whole bunch of other tricks and
years of effort from tricky designers with PhDs have gone into these new
products. It's easy for commentators like me to see clearly when we're
standing
on the shoulders of giants.
Here's the science part.
What we're seeing today in the SSD market is a renaissance
in computer architecture - and genuinely new ways of solving performance
problems. That follows 30 years of predictable developments which derived
more from foreseeable
incremental
semiconductor process technology improvements (faster, denser chips) than
unforeseeable new ways of designing digital systems.
Although
it may seem dismissive to say that nothing very new has happened in CPU design
in the past 30 years - I'm talking about architectural design. If you take as
the starting point Illiac 4
(an experimental multicore processor built in the early 1970s - which taught
architects the need for new compiler technology and fast datapaths) you are on
the straight line route which takes you to today's multi-core processors on a
single chip. Yes - there was a
RISCé wobble
in CPU thinking in the mid 1980s - with a decade of ripple effects - but once
microprocessors and memories were designed with the aid of CAD tools - rather
than taped out by hand - most of the truly creative thinking about CPU
hardware architecture went into micro rather macro level design concepts.
Today
in the SSD market - the way to design an ideal SSD for any particular
market is still a process which owes more to human ingenuity and new patents -
than to running another pass through a logic compiler.
And that takes
us to another view of the SSD market which has clarified in 2008 - the growing
importance of Intellectual Property. If you look behind the stories of who has
been suing whom in the SSD market or who has been trying to
acquire whom - you'll
soon reach the conclusion that IP in SSD controller design (and who owns it)
is as important today - as knowing which flavor of assembly language was going
to dominate the desktop PC market, or workstation market in the early 1980s.
Although
memory plays a big part in making up the cost of an SSD - it's the
controller IP which
determines the personality of the SSD - and the price that customers are
prepared to pay. The nightmare scenario for any memory chip maker is the risk
of being locked out of the SSD box - and being forced to supply a vastly
expanded market for memory chips for SSDs merely as a commodity supplier. Years
of developing partner relationships with PC and server oems could become
irrelevant - if the SSD market becomes a bigger market for their chips with new
rulers.
And going back to the western movie analogy above - another
scene is when our hero (the customer) walks into the saloon after a long ride
(having seen many wild horses - none of which match the descriptions that the
wrangler is looking for).
"Pour me a drink" says our
androgynous hero. "No, not the everyday stuff - the good stuff from behind
the bar."
That leads to the another question which has arisen this
year - and which will receive more attention from vendors and users throughout
2009.
What is the good stuff? - when it comes to SSDs? Can you
trust the strength of the juice as stated on the label? I've already
indicated that many
published SSD specs fluctuate - and many magazine benchmarks for SSDs can't be
relied on.
Looking ahead to 2009 - what can we expect? Almost
certainly - more vendors entering the market, more products, more
SSD jargon/babble -
and the start of public acrimony regarding disputed claims over market
leadership and competing visions for the way ahead. I predict we'll call 2009
- the "Year of SSD Confusion".
As the SSD industry moves to
the next stage in its growth users and oems will, like
Indiana Jones, have to learn to
tread carefully to avoid pitfalls on their way to SSD's mythical treasures.
(Trust me - I've seen the movie.) There's a serious risk of backwards steps
and also disaster striking from the beginning to the end of the SSD path.
Is it worth it? I hope so. There will be amazing outcomes for
many. But one lesson from zoology is that survival of the individual is not the
same as survival of the species. Even if the SSD market does grow to $10
billion annual revenue (as I predicted in 2003) many SSD vendors will fail.
Many promising SSD product avenues will turn out to be dead ends too.
As
always - StorageSearch.com will be
here reporting on what
happens and helping you find your way safely to where you'd like to go. | |
| . |
There
are hundreds of articles about SSDs on StorageSearch.com
The top 50
or so most popular SSD articles are summarized and updated
here. | | |
| . |
|
| |
... |
several vendors have already announced they each expect to get 50% of the
future market for SSDs |
 |
Spellerbyte's
ScryWareTM
utility often
provided spookily accurate market predictions. |
| ... | |
 |
| . |
| SSD Market Missed
Opportunities in
2008? |
Lest we get carried away on a wave of
optimism and market hype - it's important to remember that there have also
been some big missed opportunities in the SSD market - in 2008 too.
In the interests of balance I've listed them here.
- The abysmally bad design integration of flash SSDs in most notebooks -
which completely miss the true potential of this technology described in my
SSD market
model.
The result has been a wave of yawn inducing products
which you would not rush out to buy if your best friend had just bought one.
Whether this has been due to a lack of imagination by notebook product
marketers and designers - or simply the need to rush out designs which tick the
SSD box - I don't know. But I haven't yet seen an SSD notebook I would actually
buy myself. Products on offer at the moment are as inviting as a battery
operated engine in an offroad 4x4 car.
- Several years ago - the big
hard disk makers paid
out to settle class action suits related to misleading claims about capacity. I
expect that the flash SSD market will be fertile ground for future
performance-related lawsuits from oems and users. I analyzed the issues in
Can you trust flash
SSD specs & benchmarks?
From my discussions about this subject
- it seems that the root causes lie more in vendor naivety rather than
duplicity. But that's little consolation. The industry needs to get its act
together. |
| . |
| SSDs Pass
HDDs in Storage Density |
2009
may well be remembered as the year that flash
SSDs surpassed HDDs
in storage capacity in the same form factor.
I'm not talking about itsy
bitsy
1 inch and smaller drives
here. I'm talking about the hard core
2.5" form factor.
That's
the size which once seemed to offer the
best hopes
for hard disk makers staying in business - in applications like disk to
disk backup, entertainment
bulk storage etc.
In January 2009 - pureSilicon started
sampling a
2.5" MLC SSD -
with 1TB capacity in a 9.5mm high form factor.
A few weeks later
Western Digital
temporarily restored the parity in storage density when it announced a
2TB
3.5" hard drive. Since you can put 2x 2.5" drives into a single
3.5" enclosure - you
can think of them as being equivalent. That is until either the next
amplification in MLC (if it ever
works) or the next shrink in flash memory (maybe
later than sooner).
Price of the 2.5" terabyte SSD wasn't
mentioned. I expect it will cost a lot. But nowhere near as much as the 1st
terabyte SSDs cost - when they appeared in
2002 - at
a cool $2 million.
So you may well ask - when will SSDs cost less
than HDDs for the same capacity?
In some high-performance grades (15K
RPM server drives) - I expect to see that happen this year - in smaller
capacities like 100GB. Looking Ahead to the
2009 SSD Market |
| . |
 |
| . |
| PCI Express SSD Searches
Overtake 3.5" SSDs |
| Editor:- February 3, 2009 - StorageSearch.com
disclosed that in January pageviews for
PCIe SSDs had
overtaken 1.8" SSDs,
and 3.5" SSDs
too, placing them at #2 in search destination popularity behind
2.5" SSDs. |
| . |
| Notebook
SSD Market Overview - still not a pretty sight |
Editor:- June 15, 2009 - StorageSearch.com
published a new article today called -
Overview of
the Notebook SSD Market.
There's a simple way to summarize
the complex view of the SSD Notebook / Netbook market.
Lots of
initial hype and optimism that the market would deliver an astonishingly
new product experience to users, followed by dismay and disillusion due to
a flurry of poorly conceived, badly designed and ineptly executed products.
...read the
article
|
| . |
| Introducing - Fat,
Regular, Skinny SSDs |
Editor:- July 28, 2009 - StorageSearch.com today published
an article -
RAM Cache Ratios
in flash SSDs - which proposes new terms to describe and differentiate
products in the flash SSD market.
It is hoped that the new
classification jargon will be useful to users who have to evaluate lots of
products, and will be useful to vendors as a shorthand when communicating
about different segments within their flash SSD product lines.
Within the 2.5" SSD
market - for example - the
fastest products
confusingly include models in all 3 of the new categories.
The new
article explains why it's important to know the underlying RAM cache
architecture - even if you're happy with the R/W and IOPS performance. ...read the article |
|
| |
