| leading the way to the
new storage frontier||..|
what's the state of DWPD?
meet Ken and the SSD
Are you ready to
rethink enterprise RAM?
SSD endurance myths
and legends - now in 3D
pressures and projections in enterprise flash
|The SSD market during its
short history (spanning only 40 years) has managed to accrue an imaginative
body of literature which includes truths, half truths, mysticism,
misunderstandings. myths, legends - and in some cases - downright balderdash -
when it comes to the subject of SSD costs, pricing and justifications.
Astrological Age of Enterprise SSD Pricing|
|stories from SSD news in May months gone by
2012 - Buffalo puts MRAM into SSD
SanDisk acquires Pliant (a SAS SSD company)
2010 - SandForce
launches branding program
2009 - Unity Semi
says CMOx will replace nand flash
2008 - Mtron
supplies NASA first terabyte SSD in Space
2006 - Samsung
unveils world's first hybrid for notebooks
2004 - RamSan
rackmount SSD is IBM TotalStorage Proven
see more in
thank you for all the good work on storagesearch.com. I have been following it
for about the past year.
|Bob Pearson, Principal
Engineer - Cray|
(email to the
editor - March 25, 2014)
|PCIe SSDs for use in
enterprise server acceleration have been shipping in the market since 2007.
It's one of the most popular SSD subjects pursued by our readers (and has been
With over 100 million enterprise PCIe ports already shipped - the
converging PCIe SSD / server market is well positioned to expand into new
|the PCIe SSD directory on
| Skyera's history of
flash memory and storage video|
|Editor:- March 27, 2014 - I was a few weeks
late in discovering this talking heads SSD video... But I don't think 2 weeks
matters too much when the subject is the
history of flash memory
and storage (video) - and the main speaker is Frankie
Roohparvar, COO - Skyera - who's been
in the non volatile memory business for 30 years and has over 480 patents (in a
conversation with David Davis
a really non technical way (which even a
VC or lawyer can
understand) Frankie Roohparvar covers the ground from the earliest nvm (cavemen
making marks), lists the iconic use cases for flash based products like ipods
and cameras - which kept this market alive and innovating - and brings things
right up to date with the business thinking inside Skyera's
petabyte scale SSDs.
|If your education didn't include semiconductor
physics - and you're still struggling with imagining how
MLC differs from SLC and
eMLC, and how
controllers and all those electrons (locked in leaky cells) inter-relate
to each other - Frankie's verbal explanation will make it all fit into place.
On the business case for Skyera's approach - Frankie Roohparvar
reaffirmed something which I think the company has always been clear about - "When
90% of the system cost is flash you really need to understand the
internal workings of flash (to drive the cost down to a new lowest level)."
...listen to the video
amplified usable capacity|
Editor:- May 20, 2014 - When I saw
yesterday that Kaminario
would be talking about a
cost of around
$2,000 per usable terabyte of SSD storage in its new 5th generation K2
today - my gut feel was - this price metric must be based on some kind of
amplified capacity figure - rather than the conventionally
raw storage capacity - because I know that Kaminario isn't in the
deep flash IP
controller business - and unless I've been asleep and missed something -
this kind of headline figure isn't coming from arrays of
So I asked the company about it - and was told that the "average
$2/GB usable cost includes data reduction with Kaminario's guaranteed effective
capacity offering (meaning they will provide customers with incremental free
hardware if their original capacity needs aren't met)."
I probed again - does it apply to channel sales and well as direct sales - I
got confirmation that it does.
I think this is very significant.
Because guarantees are the way to add weight to an SSD vendor's
convictions that they are confident about what they're claiming.
9 years ago in the
enterprise SSD market - we saw the world's first performance guarantees
being given - about raw speed claims.
And similar to that in
significance - I think that
K2 v5 product launch will be remembered - not for the product
specifications - but for ushering in a new era in enterprise SSD marketing -
in which the increased utilization from
software - stops being
a wishy washy averaged (sometimes you might get it, but other times you won't)
concept which sounds good in a news headline - of the type we've seen from a
lot of other vendors in recent years - and instead becomes a guaranteed figure -
like endurance - which users can learn to trust.
If Kaminario can
behind its usable capacity amplification claims - then other vendors should
do the same - or stop talking about promises they're not confident enough to
And if some vendors say - we don't know the workload - so
how can we provide a guarantee? I'd say - get to know your customers better - or
risk losing a lot of business because the alternatives are offering over
specified, over priced systems - which work - or under-specified, under-priced
systems which might fail early.
is life really better with Plextor?
Editor:- May 20,
2014 - The paucity of serious marketing
ideas in the consumer
SSD market nowadays can be judged by the fact that Plextor America today
launched a marketing program - "Life Is Better With Plextor" -
which involves consumers signing up for a weekly lottery in which one person
each week can win a free SSD.
comments:- I only mentioned it here because it can be amusing to witness
the silliness of technical companies when they're engaged in consumer markets.
Nimble says 200 customers have used its bundled stack
May 19, 2014 - Nimble
that over 200 enterprise customers have used its
(pre-validated reference architecture) which is centered around the company's
hybrid array news,
PMC blog discusses latency implications of DSP ECC IP in SSD
Editor:- May 15, 2014 -
in LDPC-based Next-Generation SSD Controllers is a new blog by Stephen Bates,
Technical Director, PMC
who says - "The variability of the LDPC decode time is a function of how
many iterations it takes to decode the data from the flash."
his article Stephen says that the minimum number of iterations is 1, typical
is 4 and maximum is 20.
To relate that to latency - he says assume for
sake of illustration that each iteration takes a microsecond.
comments:- you can see how those numbers can start to stack up and make
inroads into the design of fast
one of the reasons you've got so many
different generations of
flash memory circulating in the same market today.
The higher the
capacity of the SSD - the greater the economic incentive to use newer smaller
flash geometries. But those require more complex controller management (to
integrity) so that incurs greater design complexity and NRE.
market is the enterprise - but these DSP flash concepts are used in industrial markets
too. In fact that's where they originated. Bu in industrial SSDs it can still
be sometimes cheaper
to deploy more expensive SLC memory in low capacity designs - due to the
simpler requirements of the associated controller technology and therefore also
lower demands for
power hold up
PS - in an earlier blog in this series - Stephen Bates
(whose PhD was in signal processing) - revisits the reasons why the
SSD market needs
to consider the design freedoms which come from using complex DSP flash IP -
gives examples of the tradeoffs. Such as 50% better
with LDPC codes using identical flash - or gaining
capacity by using weaker codes.
BTW the industry changing
possibilities of these technologies for reshaping the economies of SSDs were
reviewed in my 2012 article -
flash care management & DSP IP in SSDs What is it? Who does it? and why?
Oracle acquires GreenBytes
Editor:- May 15, 2014 -
it will acquire GreenBytes.
also:- SSD software,
companies from 2000 to the present day
Virtium promises 4 years "no requals"
May 14, 2014 -Virtium
that its new 2nd generation
(SATA 3 compatible) can deliver upto 4x the read performance of its 1st
They're available in in 2.5", 1.8", M.2,
mSATA, Slim SATA, and CFast form factors.
Commenting on the high
amortized cost per
unit of requalifying SSDs in embedded industrial markets Scott Phillips,
director of marketing at Virtium (who
joined the company from HGST)
said about the issue - "With its 2nd generation StorFly SATA SSDs,
Virtium is able to guarantee that its SLC-based StorFly PE class products will
not cause a requal for at least 4 years."
Because Virtium did such a good job a few years ago explaining to me how it was
adapting its technology roadmaps to deal with MLC - I mistakenly assumed for a
while that all the StorFly SSDs were only available with MLC inside. But I was
wrong. The new models are available with either MLC or SLC.|
new report on the costs to 3D nand flash manufacturability
May 14, 2014 - Sometimes a good question can reveal as much about the state
of the SSD market as the answer.
There isn't always a pre-packaged
answer to many of the questions I get from thoughtful readers. And investing
the time to understand why that is so - when the question is good one - and
there should be an answer often leads to me writing new articles.
you'll be pleased to know that there is a packaged answer to this
question posed by Gregory Wong,
Insights who asks...
"What is the incremental investment
required to transition a 32 layer 3D NAND fab to 64 layers? What is the impact
on fab cycle time and manufacturing capacity?"
But you'll have to
buy his latest report -
Investment Implications of 3D NAND (overview pdf) - to find out what it
Forward Insights - SSD
and nvm reports overview
is it time for enterprise SSD designers to reconsider RapidIO?
May 14, 2014 - You'd think that with all the interfaces
already in use
within the enterprise SSD
market - there wouldn't be enough of a market gap to justify introducing
yet another one. - Particularly when that interface strays across low
latency server-storage territory which is dominated by
PCIe SSDs, under
attack by memory
channel SSDs and has been flanked historically by
thought so too.
But in a recent article -
You Really Know RapidIO? - by Eric Esteve , founder of
IPnest says - "Maybe it's
time for the server/storage industry to give a second chance for the RapidIO
Editor's comments:- That's a bold statement -
coming as it does from someone who was involved in designing one of the first
generation PCIe controllers 10 years ago. Eric argues that the intrinsic fabric
architecture and routing support in RapidIO - would make many of the things
which architects are trying to do today - such as interconnecting large numbers
of servers and SSDs for example - easier and faster.
SanDisk and Toshiba collaborate on 3D nand fab
May 13, 2014 - SanDisk
and Toshiba today
that they have begun work on demolishing and converting a 2D NAND fab at
Yokkaichi Operations, in Mie prefecture, Japan over to 3D capability with a
view to enabling 3D output in 2016. See also:-
a new CMO for Fusion-io
Editor:- May 12, 2014 -
Mendenhall (who from 2007 to 2011 was CMO at
HP) has joined Fusion-io
as executive VP and CMO.
Editor's comments:- this fills a long
standing vacuum in FIO's marketing - initially created by the
management changes a
Being CMO at Fusion-io is one of the most important
marketing positions in the SSD industry. The whole industry looks towards
Fusion-io as an indicator of future directions in the enterprise - even its
TrendFocus reports SSD shipments in Q1 2014
May 10, 2014 - According to a
report by TrendFocus
- the top 2 SSD companies (based on the number of drives shipped in Q1 2014)
with 32% of the market and 26% respectively.
You can see a table
listing the other top companies in the
summary by StorageNewsletter.com here.
Samsung starts 3D nand production at new fab in China
May 9, 2014 - Samsung
that its new memory fabrication line in Xi'an China - which will make 3D
V-NAND - has begun full-scale manufacturing operations.
"SATA inside" SanDisk's ULLtraDIMMs
May 8, 2014 - Today I learned some new details about the architecture in SanDisk's
SSDs by watching a video -
Architecture and Design Technology - which was recorded at the recent
Storage Field Day event
(April 25, 2014).
Inside each flash DIMM are 2x SATA CloudSpeed SSDs
which connect to the DDR3 interface via a Diablo designed bridge
chip which has something on the order of "megabytes" of buffer memory
That gives a much better basis for
performance limits of these devices.
And it confirms why you only
start to see noteworthy performance from them after installing 3, 4 or
- despite the "SATA inside" limited nature of this 1st generation
memory channel storage design - the performance - aggregated at a block level -
results in a usable write latency envelope which
IBM has verified to be in
latency category than most leading
- if you're thinking long term - and roadmaps - it's reasonable to assume that
if a 2nd or 3rd generation of MCS flash DIMM SSD could abandon the "SATA
inside" and use instead a
architecture flash controller with a native memory array optimized type of
bridge interface - then the gaps between what MCS can offer - compared to PCIe
latency - could widen.
By the way in this 52 minute video I also
learned 2 interesting details about
SanDisk's CloudSpeed SSDs
which I didn't know before:-
SSD news and articles
- SanDisk's Guardian technology doesn't perform scheduled data integrity
checks to detect silent errors for data at rest.
The stated reason
being that this would eat into
And the implied assumption being that SanDisk thinks this type of activity is
best determined by the system designer. So - if you design these SATA SSDs
into a storage array - then you'll have to take responsibility for that aspect
of long term data health in your own software.
Intel invests in Maxta
Editor:- May 7, 2014 - Intel
Capital was one of the leading investors in a $25 million series B investment
in Maxta which
today. Other investors included Tenaya Capital and Andreessen Horowitz.
comments:- these investors aren't the only ones who believe Yoram Novick's
claim that the company he founded is markedly different to the other 95% or so
of SSD software startups.
A lot of you were sufficiently curious too - which elevated Maxta into the
Top SSD Companies list
in Q1 2014.
Maxta case studies and white papers,
new trends in
SSD boxes, VCs in SSDs
Avago completes acquisition of LSI
Editor:- May 6,
2014 - Avago Technologies Limited
it has completed its acquisition of LSI an all-cash
transaction valued at approximately $6.6 billion.
creates a highly diversified semiconductor market leader with approximately $5
billion in projected annual revenues.
Avago believes the acquisition of LSI repositions Avago as a leader in
the enterprise storage market. The acquisition also expands Avago's product
offerings and brings system-level expertise in its wired infrastructure market.
LSI will operate as a wholly-owned indirect subsidiary of Avago and
will continue to conduct business under the LSI name.
comments:- A quick way to graps what Avago is all about is to view their
LSI's SSDs and controllers already operate in more markets than you can easily
count - it's interesting to speculate what integrating Avago's connectivity
technologies with LSI's enterprise flash controller SSD expertise will do
for the SSD market.
My guess is that it will become easier, cheaper
and less risky for systems designers to place SSD functionality anywhere
they like without having to dive deeply into the interface details.
EMC acquires memory channel SSD in a box company DSSD
May 5, 2014 - EMC
has acquired a stealth mode rackmount SSD company - DSSD
- it was announced
Products based on the new DSSD architecture are expected to be
available in 2015.
Editor's comments:- an informative and
entertaining article about this can be seen on
sounds like the DSSD product will implement a large directly addressable (by
PCIe) memory space - based on a transparent
RAM cache flash
scheme and will offer latencies similar to
SSDs - but with the capacity difference being that you can pack more
capacity into a box than in a bunch of DDR3 DIMMs.
It hearkens back to
the original big shared memory in a box connected by PCIe of
Violin's first product
- the Violin 1010 Memory
Appliance - which was launched back in 2007.
Although that was
pure RAM - and
enterprise users in those days weren't as
educated as they
are today. That meant Violin had to go back and redesign the product to include
a fibre-channel front end -
to make it fit in with the market idea of what an enterprise SSD box should
really look like. And that product also came in at the tail end of the
RAM SSD market - in the
year before new flash
controller architectures enabled SSD makers (including Violin) to displace
low to mid range SSD boxes with pure flash.
And before Violin's product
- in 1994 - Texas Memory
Systems was shipping a product called the SAM-2000 (Shared Access Memory) -
which enabled a bunch of different computers (even with different OS and
internal busses) to share the same memory at low microsend latencies and bus
There's an argument for saying that Violin already offers
something similar now to what EMC hopes to ship next year - in the shape of
Flash Array (WFA). Except that the 1st generation WFA uses GbE as the server
clustering fabric. But my guess is that it would be easy for Violin to offer
other variations with lower internal latency - to connect the CPUs to its
- if they thought enough customers would buy it.
IBM can already do
something similar (to Violin) with its X6 architecture.
limitations in IBM's technology are that the 1st generation of eXFlash DIMMs
(designed by SanDisk
and Diablo) are greedy
when it comes to power consumption. That means legacy server motherboards
don't have enough current capacity routed to the DIMMs to enable you to fill
them all with these modules. That's easy to change, however by adding more
copper in the motherboards and increasing the air flow in the box.
EMC offers the DSSD as a product which is unbundled from the brand of server -
then it might find a niche market for it.
But there are so many
different ways of tackling the same problem (such as enabling
PCIe SSD fabric from
storage, not to mention what's the PCIe interface for in
Skyera's new box)
that it may already be a crowded market by the time EMC has anything to ship.
Are you ready to
rethink enterprise RAM?
controllers inside EMC's DSSD
- key changes in the SSD market,
history - all|
| new whitepaper from
Diablo looks at the "drawbacks of traditional PCIe SSDs"|
|Editor:- April 22, 2014 - Diablo Technologies
today published a white paper
Storage Performance: it's all about the Architecture (and not just the
|This paper presents Diablo's perspective - as
the creator of
Storage (an architecture which supports terabyte class flash SSDs in
unmodifed DDR-3 DRAM DIMM sockets and runs within those safe operating
electrical power envelopes).|
The article looks at the measured
performance weaknesses of some specific (but unnamed)
PCIe SSDs with respect
to their measured performance
under various loads and then attempts to generalize those weaknesses as part
of an argument that MCS is a better solution than PCIe SSDs - because MCS is
more scalabe and doesn't suffer from the same kind of architectural
comments:- This paper from Diablo reminds me of the type of comparisons
between different types of PCIe SSDs (and their sensitivities to different
workloads) which was one of the pivotal marketing points of difference which
Virident used to
hang their reputation on in the years leading up to its acquisition by
HGST. (Although as I
often reminded readers at the time - Virident wasn't the only company with that
kind of array scalability or no-compromise performance.)
Going back to
Diablo's white paper - for me - like many vendors written papers - it's good and
bad in different parts.
The best bit is the middle - in which you get
a reminder - from measured results that fast SSDs with similar capacities
The worst part in my view is the attempt to link
these comparison results to a general conclusion about the merits of MCS
versus PCIe SSDs.
Because I think the architecture of the
controllers on the
flash side of the SSDs plays such an important part as too does the
software. And that's
still just a small part of the picture.
My considered view is that
Diablo's extrapolation towards a general market conclusion from one
selected comparison example is not merited by the evidence presented in this
And while I am convinced of the benefits of Diablo's MCS
architecture - and what it can do differently and better compared to PCIe SSDs
- I think this particular paper isn't the best case they could argue.
|"...My advice re
SSDs for database acceleration has always been - try before you buy.
That's because the performance model which you have in your head may not be the
same performance model which is at work inside your system."|
|...Editor (in many conversations) about the
interplay of enterprise software with SSDs in database apps. |
sudden power loss|
|Why should you care
what happens in an SSD when the power goes down? |
This important design
feature - which barely rates a mention in most SSD datasheets and press releases
- has a strong impact on
SSD data integrity
This article will help you understand why some
SSDs which (work perfectly well in one type of application) might fail in
others... even when the changes in the operational environment appear to be
|This article proposes a
simple scheme which enables the essential characteristics of an SSD enhanced
server to be communicated by a single number which tells you all you need to
know about the latency and likely performance from an SDS viewpoint.|
|SSDserver rank 0 to 7|
petabytes shipment snapshot of enterprise flash|
|Editor:- May 9, 2014 - A
recent blog - in ArchitectingIT
- says that 3 leading vendors shipped a sum total of over 50PB of
rackmount SSDs in
Q1 2014 - with HDS
- apparently having shipped more flash capacity than either EMC or Pure Storage
according to estimates by the blog's author Chris Evans.|
comments:- How does this compare to other vendors? and to other times?
See Petabyte SSD
Milestones from Storage History.
Another context is this.
and when Skyera
ships just 40U of its fully populated upcoming
skyEagle (in a single
quarter later this year) that would be more flash capacity than anyone in the
above list. (Although they should all be shipping more too if you believe
flash capacity - on its own - is a crude and meaningless measure. (Unless -
I suppose - you're the company which sold the
chips or the boxes.)
of enterprise flash depends on how fast it is,
where it's located in the
datacenter architecture and how well it has been integrated to leverage
the application architecture.