product line - from
WD Solid State
Storage (formerly - SiliconSystems)|
started talking to me publicly in 2004 they communicated an obsession
with reliability and their enthusiasm for a new business model in which their
flash SSDs would replace hard drives in embedded appplications for an entirely
- not because they were rugged - (in all these applications
HDDs had been used
- not because they were faster - (although they were faster - the
company wasn't chasing the speed tiger like most other SSD makers)
- not because they were cheaper - (the SiliconDrives initially cost
more than the HDDs they were expected to replace)
Another unusual thing at
the time for a flash SSD company focusing so much on
that SiliconSystems wasn't in the
- but because they would be cheaper to own in applications in which
users looked at the total cost of maintaining large populations of systems which
had high service cost.
Perhaps because the systems were
geographically dispersed (unlike in a datacenter). Users could start to count on
year on year cost reductions due to the theoretically (at that time) higher
reliability design of the new SiliconDrives.
Until that time most flash SSD companies had previously been in the defense
markets and had served long apprenticeships to prove their reliability
Was the SiliconDrive really going to be part of a
genuinely new market segment in the SSD market - or just wishful thinking?
By the time I published the 2nd edition of my
penetration model (in 2005) I was sufficiently convinced to add it as one
of the SSD market segments.
Before the SiliconDrive it was unusual for
SSD makers to talk in much detail publicly about the architectures of their
products which they regarded as commercially sensitive proprietary information
protected by patents and NDAs.
This was something I knew from my own
dialogs with flash SSD oems and their reluctance to talk publicly about
important issues like
Military SSD companies did talk in detail to prospective designers and
preferred to resolve issues related to reliability behind closed doors.
from the outset SiliconSystems's online communications and web advertising
opened the door to a view of what was happening inside a flash SSD. They
talked frankly about reliability issues with flash memory and how their
architectures managed those problems.
For example in April 2005 the
company published an article here on StorageSearch.com -
SSD Reliability - which showed why their way of doing wear leveling could
achieve much higher operating life than other similar schemes.
context of SSD
market history - 2005 was the year when
SSDs were a strategic market - the 1st time that a multibillion dollar company
had entered the market. In the years which followed the hiatus into the
market - it became clear to readers that there were 3 things which remained
different about the SiliconDrives - compared to the many other
2.5" SSDs in the
commercial SSD market.
- SiliconDrives weren't very fast. In fact they were often 2x, 3x or
4x slower than the fastest SSDs in the same interface / capacity class.
- SiliconDrives didn't offer high capacity. The maximum capacity
of SiliconDrives was often 2x and sometimes 4x less than other 2.5"
When I challenged the company's marketers about these points -
far from being concerned about looking less competitive than other drives in the
market - they told me there were good reasons for them never appearing in the
lists of the fastest
or highest capacity 2.5" flash SSDs. To summarize what they said and what I
was able to interpret for myself - the arguments went something like this.
- SiliconDrives were exclusively SLC. This was an
- speed:- the applications for SiliconDrive didn't need ultimate
speed. One trade-off in an
SSD controller for a
SATA interface is that
if you dedicate a lot of CPU resources to issues like
active wear leveling, garbage
collection etc - then the only way to increase performance too is to put in
a faster and higher powered processor. That affects the power budget and
reliability of the SSD - and could move it outside the zone where it is so
attractive for embedded applications.
- capacity:- most of the applications for early SiliconDrives were
replacing HDDs in embedded systems. Unlike the notebook and enterprise server
markets - these were on a different capacity roadmap. And most legacy apps
didn't need headline capacity drives.
In the years between 2005 and 2009 SiliconSystems
published a host of
white papers on flash SSD design issues - partly to make customers realize
that serious reliability issues existed in all flash SSDs - and mostly to show
that the company had a good story to tell about how these were managed inside
- SLC vs MLC:- the answer here was unambiguous. The designers of the
SiliconDrive did not think MLC was reliable enough for the reliability envelope
they were aiming for. MLC would not be appearing in a SiliconDrive any time
I wondered about that point - because I knew the market
pressures that 2.5" SSD makers were under. I expected that one day - the
company which made the SiliconDrive would also want to engage in the MLC SSD
market. I knew that would cause difficulties in branding.
would a customer be expected to differentiate between different grades of
reliability in a reliable SSD?
That decision still lay some
years in the future.
The reliability messages went outside the SSD too
- when the company alerted designers to the importance of laying out their
boards carefully when using PATA drives to
electrical noise - and the importance of building power sequencing testers
to detect weaknesses in competing SSDs - which might not show up in most steady
Was there really a market for this type of
commercial SSD - which wasn't following the fashions of the rest of the
The market got an answer to that when the world's 2nd largest
hard dsk maker Western
Digital acquired SiliconSystems for $65 million in March 2009.
A few quarters later
when I spoke to Gary Drossel, VP of Product Planning, in WD's SSD business unit
he emphasized how big was the investment made for long term testing. He joked
that the large number of their SSDs now undergoing long term tests in WD's labs
would have almost made the Test Labs one of SiliconSystems' top 10 customers not
so long before.
Then the inevitable happened. In March 2010 -
the new owners of the SiliconDrive brand
entered the consumer
SSD market with an MLC based SSD.
What was not so inevitable was
that instead of diluting all the value acquired through 6 years of promoting
SiliconDrive as synonymous with "reliable SSD" - they called the new
MLC family SiliconEdge
That was a very astute marketing decision.
Reliability is indivisible. You've either got it or you haven't.
the SiliconDrives - now in their 4th generation are starting to get a little
WD had no idea I was writing
this article - and as with all the market analysis articles I write - vendors
mentioned in the text only get to see the articles after they have been
published and are visible to all our readers. I always welcome insightful
comments which can be appended to any new articles in the days immediately
following their 1st appearance.
Here are some more articles on the
theme of SSD reliability
- SSD recovery
- one way of segmenting the SSD market is to categorize SSDs into products from
which you can recover data if the SSD is damaged or faulty - and those from
which you can't. This issue matters a lot for some types of users and
applications whereas it's irrelevant for others. There are some technical
gray areas - but there are hard boundaries too. As
everything which goes
into an SSD has a cost - I'm surprised that "data recovery supported
design" isn't a bullet point in
- SSD power
down management architectures - first published in February 2011 - this
ground breaking industry survey of a topic rarely mentioned in SSD product
datasheets could spark as much interest in this subject in coming years as our
classic endurance article did when it first hit the web.
ads helped to build the reliability message|
|Here are some examples of banner ads
for the SiliconDrive product family which ran continuously here on
StorageSearch.com in the period from 2005 to 2012.|
As mentioned in
the case study notes on the left - these were a small part of the web marketing
assets which also included other types of ads, press releases and articles.
are some gaps in the series shown below - when some designs were changed to
The purpose of showing them here is not to
illustrate how to design
high click rate banner
ads - but to show how a succession of consistent messages evolved over a
period of time helped to create a trusted brand in the SSD market in the
sensitive segment of SSD reliability.
The banner ads were mostly used
to educate potential customers -by linking to white papers which
described the technologies and concepts
in the banner messages.
That's in contrast to the more common use of
web ads (generally) which is to link straight to products which a customer can
The design and vendor qualification process for high reliability
SSDs in embedded systems takes many months - and it takes longer for new
companies because they have to prove to prospective customers that they really
know what they're talking about and they will still be around in the long term.
simplicity in the context of this branding article I've linked the banners below
all to the same url which is a collection of white papers.
banner ads ran originally appeared on this site - they each linked to
individual papers or related content - some of which has since moved.
|start date March 2005|
|start date November 2005|
|start date June 2006|
|start date March 2008|
|start date January 2009|
|start date July 2009 - end date August 2012|
is published by