See also:-
LSI
- editor mentions on StorageSearch.com and
LSI's
SSD page |
|
. |
Editor:- April 30, 2014 - LSI was ranked #3
in the Q1 2014
edition of the
Top SSD Companies List
which is researched and published by StorageSearch.com |
|
Who's who in SSD? - LSI
by
Zsolt Kerekes,
editor - November 18,
2013
For the past year or so I had been wondering if the glory days
of LSI's SSD controller technology lay mostly in the past. But I can now
understand why it took them so long to complete this new design - which is
almost at the integration level of "SSD market in a chip". |
|
Editor's earlier comments:- October
2013 -
LSI - ranked
#14 in
the top SSD companies in
Q3 2013 - operates in the
PCIe SSD,
SSD controller and
auto tiering / SSD ASAPs
markets.
The company sampled its 1st SSD product in
March 2010.
It was a
PCIe SSD - which is
software compatible with SAS
- an interface which LSI
helped to pioneer. Elements of the IP in the new SSD design came from
Seagate and
SandForce.
LSI
acquired SandForce in
January 2012 -
and revealed the strategic significance of this shortly afterwards when it
revealed that LSI's PCIe SSD product - the
WarpDrive
- was being oemed by EMC.
LSI's SSDs are
small (rather
than big) SSD architecture and span a wide range of market applications
including:- enterprise
SSDs, industrial
SSDs and consumer
SSDs.
LSI's interim approach to
adaptive DSP
ECC IP for SSDs - revealed in August 2013 - is to use tactical LDPC - as a
try it again strategy for improving data usability. |
. |
|
In March 2009 -
LSI announced better
support for flash SSDs
in the latest update to its
MegaRAID
SAS adapters. LSI calls this new feature SSD Guard - which can anticipate
some types of flash SSD failures in
RAID 0 configurations
and starts rebuilding data on a spare unit.
In December
2009 - LSI
announced it is
sampling
the LSISAS2208 dual-core 6Gb/s
SAS
RAID-on-Chip
IC to OEM customers. It's
intended to support the forthcoming PCIe 3.0 specification, currently under
development and provide performance levels that meet the needs of
next-generation server platforms based on
flash SSD storage.
The new LSI SAS ROC will deliver performance levels of up to 600,000 IOPS.
In January
2010 -
LSI and
Seagate
announced
they have collaborated on designing
PCIe SSDs for the
enterprise accelerator market - which started sampling in March 2010.
At that time - LSI was approximately the 163rd
company to enter the
SSD market (not counting
SSD SoC makers - which
would push the score to about 185).
in November, 2010 -
Demartek
published a sponsored
test
report (pdf) which compares the performance of
SSDs and
HDDs in a simulated
web server environment when managed by LSI's
CacheCade
software - which provides
SSD ASAP
functionality.
Editor's commnents:- The report shows that
throughput and access times were improved by at least 3x using a single
SSD cache compared to the HDD only situation.
However - it's
disappointing that the sizing of the test was not best chosen to draw
meaningful conclusions. Because the web content was only 25% larger than
the SSD capacity! It would have been more helpful to design a simulated case in
which there was at least a 10x or 100x size difference. Because if you
can fit all the web content onto an SSD then you don't need the burden of the
"cache" software at all - and might get better results by switching it
off.
There are case studies going back nearly 10 years which show that
SSDs can provide big speedups in web servers. The exact speedup depends on how
fast the SSD is. This test report doesn't answer the question - is LSI's
CacheCade useful in a realistically scaled environment?
In March 2011 -
LSI finally spun off the
Engenio systems
business - selling it for $480 million to
Network Appliance.
In June
2011 -
LSI was one of
several compatible companies named in
FlashSoft's launch of
its auto tiering SSD
software.
In October 2011 -
LSI
announced
a definitive agreement to acquire
SandForce for
approximately $322 million. The transaction is expected to close early in the
first quarter of 2012. SandForce president and CEO, Michael Raam will
become General Manager of LSI's newly formed Flash Components Division.
In
January 2012
- LSI
announced
it has completed the acquisition of
SandForce. And LSI
also announced
that its PCIe SSD product - the
WarpDrive
- will be oemed by EMC.
In April
2012 - LSI
announced
details of its new
Nytro
family of SSD technologies - which integrate and join up several previously
standalone elements in its product line in a new unified marketing roadmap.
In
June 2012
- LSI
demonstrated its SandForce
SF-2000 flash controllers working with Toshiba 19nm and Intel 20nm NAND
flash memory at Computex 2012
in Taipei, Taiwan.
In June 2013 -
LSI demonstrated
new features
in their
SF-2200
controllers aimed at the
notebook SSD
market. These are:- Opal compliant
encryption and
DevSleep technology.
DEVice
SLeeP (pdf) a very low power sleep mode for
SATA SSDs.
In
November 2013
- LSI launched
its
3rd generation SandForce SSD
controller family - the SF3700 - which offered native SATA or gen 2
PCIe interfaces - and incorporated adaptive R/W DSP ECC management.
StorageSearch.com described the
underlying business concept as "SSD market on a chip" - and
described the design as the most ambitious design of a single chip
SSD controller in
SSD market
history.
In December 2013 -
LSI
announced
that it had agreed to be
acquired by Avago Technologies Limited in an all-cash
transaction valued at $6.6 billion.
In May 2014 - Seagate agreed to acquire
LSI's flash
business for $450 million. |
. |
|
. |
LSI's
Nytro - positioning update |
Editor:- June 26, 2013 - I had a useful
conversation last week with Rob Callaghan who
manages outbound marketing in LSI's Accelerated
Solutions Division.
When I say "useful" - I mean - useful
to me - because I haven't spent much time looking at the details of
LSI's
family of PCIe SSD products - due to my impression that it won't tell me
anything new about the long term architectural direction of this class of
product. Having said that LSI's PCIe flash products do provide a relatively
pain-free entry point for many users into the world of enterprise SSD storage -
and will generate lots of revenue for the company in the short to near term.
Not too bad for a company which I described as "approximately the
163rd company to enter the SSD market" - when LSI announced its entry into
SSDs in January 2010.
In that same 2010 news story - I used the headline - "LSI will
Compete with Fusion-io" - because it was a useful shortcut to my
guess-ahead at what LSI might end up doing. But with the benefit of hindsight
- that headline statement isn't strictly true. LSI doesn't compete at all with
Fusion-io on a slot by slot level. And at root is the growing fragmentation
within the PCIe SSD market itself.
I started to warn about this 2 years
ago in an article - don't
all PCIe SSDs look pretty much the same? And so the fragmentation within
PCIe flash products due to the
different roles these
products can play within the enterprise - was an obvious place to start my
conversation with Rob Callaghan as LSI already has 3 functionally distinct
types of products within its Nytro PCIe flash card product line.
Having
said that - there are another 3-4 classes of PCIe SSDs for which LSI doesn't
have any products at all.
It's a no-brainer to expect that LSI will
launch products to fill some of these other gaps too. But it's also clear to me
that due to the LSI's preferences and its business comfort zones with
controllers,
leveraging its RAID and
HDD legacy and
the need to provide a simple
SSD educational
interface to channel partners - the company isn't capable or doesn't have the
inclination to go after the same kind of high end
dark matter user apps
which other vendors like Fusion-io
and Virident etc
can reach - and even if LSI wanted to - they can't get there - because of
differences in philosophy in the
sizing of their
SSD controllers and assumptions about the
SSD software operating
environment which are conditioned by
fundamentally different
beliefs about the co-existence of SSD and HDD in the future datacenter.
Anyway
- back to my conversation with Rob Callaghan who had been talking to streams
of editors etc about LSI's upcoming product
announcement
(June 24) about more new Nytro stuff. I learned 3 things.
- the new thing - LSI's Nytro Elastic Cache. One part of flash can be
configured as read cache while another part can be configured for write cache
and for semi-permanent storage of often used data such as golden images of
desktops in VDIs. As a reliability
feature - the write cache and permanent data is mirrored on 2 separate regions
of flash on the same card.
I said - that could significantly reduce
the amount of flash capacity you need on the card - while offering some of the
reliability benefits of mirroring - because you don't need to mirror data which
you can easily read again.
Rob said - you're the first person I've
spoken to who got that. (The people I speak to in these interviews know how
to schmooze editors.)
I
said - there are so many ways of getting
SSD efficiency -
each time a company does something like this it means they can make money in the
market at a price point where a competitor - using more flash to solve the same
problem - is just bleeding their
VCs.
- the thing which LSI's PCIe SSDs don't do yet. I asked if LSI
supports low latency automatic mirroring - clustering across multiple PCIe SSD
cards - like some of the high-end products from other vendors do.
Rob
said no - they don't. It will come in future LSI products - and Rob hinted
that the way they do it may use a different physical mechanism than the method
I would have expected. (For me - the norm is PCIe chip level failover
supported by PLX).
- the thing I already thought I knew. The conversation confirmed to
me my earlier assumptions about LSI's positioning in the PCIe SSD market - and
limitations of their products. (For example LSI's Nytro WarpDrive doesn't have
scalability
symmetry.)
Rob said the typical end user of this product would
mostly have just a single LSI PCIe SSD card installed in a server. It would be
rare for them to have 2. (Editor - So if you want an environment in which
every slot can be a PCIe SSD - then you're looking at a different class of
product, and different software and a different bunch of vendors.) Editor's
comments:- As I said above - my gut feel is that LSI will do well in the
market with revenues from this generation of enterprise PCIe SSDs - in user
systems which can be satisfied by fast-enough SSD acceleration in HDD heavy
environments. But the company will have to change cars - rather than just switch
lanes in the future with both its controller architecture and software base -
if it wants to provide efficient solutions for SSD-centric datacenter
installations. My guess is - if that does happen - the only viable route would
be another significant
SSD company
acquisition - maybe in 2 years time.
For a company - like LSI -
which has already demonstrated that it can successfully leverage a
significant SSD IP acquisition - which it did with
SandForce - at
technical, marketing and business levels - that shouldn't present a fundamental
problem. But until it happens you're going to see a lot of confusingly different
but parallel courses being taken by vendors in this market - which aren't
inter-compatible.
And that will lead to the
paradoxical
oddity that maybe 2 of the top 3 companies in the
PCIe SSD market don't
really compete with each other at all. | | |
. |
LSI's new
notebook technologies and the petabyte SSD shelf |
Editor:- June 4, 2013 - LSI did a really
good job leveraging their
acquisition of
SandForce.
That's
the impression I got when I was talking last week to Kent Smith, Sr. Director of
Product Marketing who wanted to talk about the
new features
in their
SF-2200
controllers aimed at the
notebook SSD
market. These are:- Opal compliant
encryption and
DevSleep technology (a very low power sleep mode for
SATA SSDs which I
wrote about in an earlier news story lower down this page).
The data
I latched onto in LSI's presentation was that according to
3rd party analyst reports
LSI's SSD controllers were used in approximately 1/3 of all the flash memory
deployed in SSDs in the client and enterprise markets in 2012.
LSI
says that even before the start of the current quarter - the SSD world had
consumed over 21 million SF
controllers.
Now the likely productivity advantages of enterprises using SSD based
notebooks were already known
even before
the SSD notebook market began. And the benefits of having encrypted drives
to reduce the cost of exposure to data loss - when a pc goes astray in an
airport or is stolen - are exactly the same for SSDs as they were for hard
drives - and therefore need no repetition here.
The intriguing
thing for me about LSI's new consumer market controllers - as I said to
Kent - was the new possibilities that they could open up in really high
capacity enterprise arrays used in solid state archives which will eventually
replace disk backup and VTLs.
In
my roadmap to the
petabyte SSD article (March 2010) I observed that one of the missing IPs in
the SSD market (at that time) which would be needed to implement petabyte scale
physical flash storage in 1U or less - was fast boot SSDs with very low sleep
power consumption.
That will enable bulk storage SSD architects to
pack an SSD array into the smallest possible physical volume - and leverage a
tape library type access
architecture at a lower cost
of ownership than tape or hard disk. But to be useful in a solid state world
- the worst case access time would need to be much faster than the 1 to 2
seconds which was the power on ready time for flash SSDs at the time of writing
that article.
LSI's implementation of DevSleep already gives a 400x
power reduction in the not needed mode - and Kent told me their power up ready
time is about 250 milli-seconds. In my view that's a good enough figure for
software architects to start planning around - and it doesn't take much of a
stretch to see how that may evolve to get shorter in the next couple of years
- if the market puts a premium on this feature.
The target for the
bulk storage SSD should be to get 5 to 10 petabytes of virtualized flash into
1U of rack height - at a power consumption level which means that every shelf
can have an identical storage density. I had a gut feel it should be do-able -
and an attractive market proposition in the 2016 to 2020 timeframe - based on
the model I published back in 2010. In some ways it should be easier now -
because I didn't anticipate just how good the
reliability
technologies for MLC and TLC would get - in particular due to the benefits
of
adaptive R/W.
So - if you have shares in a company which makes tape libraries or
disk to disk backup - you've already had many years advance warning that those
products will cease to be commercially attractive when the solid state library
market gets going. | | |
. |
we're #2 in PCIe SSDs and
growing fast - says LSI |
Editor:- May 15, 2013 - LSI today
announced
it shipped over 40,000 PCIe
SSDs in the past 12 months - and has been ranked the #2 merchant supplier
of enterprise PCIe SSDs in the US, and is the fastest growing vendor in this
category according to a recent report by Forward Insights. | | |
. |
new LSI blog on the value
of enterprise flash |
Editor:- March 14, 2013 - You
won't be surprised to see me mentioning a
recently
published blog by Robert Ober,
System and Processor Architect, LSI - about the
value of PCIe SSDs in
big datacenters - which includes these statements:-
- "Work/$ is the correct metric (and not crazy expensive $/bit)."
- "when users say - $8k PCIe card in a $4k server really? - I am
always stunned by this"
I'm guessing that the title of Robert's
blog - What are the driving forces behind going diskless? Will 100% flash
storage make sense in enterprise? - was either inspired by
SEO considerations
(stuffing the title with value-loaded words for search-engines) or was
predetermined before the blog was written.
I prefer this alternative
title - suggested by a banner graphic in the blog itself -
An $8K PCIe
card in an $4k server - huh!?! | | |
. |
NetApp and LSI do that "my
software loves your SSD" thing |
Editor:- March 6, 2013 - LSI today announced
that its
Nytro
WarpDrive (PCIe SSDs)
have been validated for use with NetApp's
Flash
Accel (SSD ASAP)
software.
Editor's comments:- According to
NetApp's pdf
- "Flash Accel has the ability to keep the cache warm and coherent in
the event of disruptive operations and restart caching from the reboot/crash
point, rather than restarting from a cold cache."
But it's not as
unique in these respects as their document would have you believe - although
this suggestion is probably because of when the document was written.
LSI
says in its press release that its "advanced off-loaded
multiprocessor architecture uses up to 4x less CPU and memory resources
than competing solutions".
Now when you see that phrase - off-loaded
- in this kind of context - you can be sure that it's a dig at
Fusion-io.
The pros and cons - in architectural
efficiency and
performance -
aren't as straightforward as they appear from this subliminal value-loaded
phrasing. I discussed these issues a few years ago in an article in
FIO's product page here
on the mouse site.
The motivational reasons you might choose LSI
rather than FIO (or the other way around) probably have less to do with whether
you understand or like the way they design
SSD controllers
(which are evidence rather than motivations of what lies behind their
SSD architecture
thinking) and instead I think the reasons you might prefer one or other as a
strategic supplier would have rather more to do with whether you're
comfortable with their different
philosophies about the best routes to the
future
of enterprise storage and, in particular, whether you agree with their
speculation of what the
destination looks like.
If you're going to be in the same waggon
train for 2-3 years - bumping along comfortably together is what's
important. | | |
. |
Over-provisioning flash
capacity in SSDs - article by LSI |
Editor:- January 8, 2013 -
Understanding
SSD over-provisioning - is the title of a new article published in EDN and written by Kent Smith, Sr. Director of
Product Marketing at the SSD controller part of LSI.
Kent's
article describes the trade-offs between performance, the percentage of
over-provisioned flash capacity and the useful impact of compressible data -
which inside SandForce controllers is leveraged to create additional
over-provisioning. The interaction between write amplification counter-measures
and the benefits of using TRIM commands on performance are also noted. ...read
the article
Editor's comments:- there wasn't anything
new for me in this article - which covers similar ground to my 2011 article -
flash SSD
capacity - the iceberg syndrome - which shows how SSD makers leverage
capacity to tweak reliability and performance.
But - having said that -
I learned about over-provisioning by 10 years of talking about it - with
many SSD companies. And some of the things I put in my own article had been
gleaned from past conversations with Kent Smith himself when he was at
SandForce - as well as various other people in
Violin,
Texas Memory Systems
and Adtron.
I'm
guessing that what Kent would have liked to say on OP may have been "trimmed"
by a word count limit in his latest EDN article.
So here are some
other suggestions for more substantial and ideas packed articles I recommend
- which Kent Smith has written in the past for other publications, and which
cover SSD controllers from other angles:-
| | |
. |
LSI ships 1
million SandForce controllers / month |
Editor:- July 31, 2012 - LSI has announced
enhanced support for the
Ultrabook SSDs
market in its SandForce
SF-2200/2100 controllers:- enabling lower SSD power consumption, faster
boot and support for "virtually all MLC flash product families".
"LSI
has shipped well over 10 million
SandForce processors
and we anticipate our shipment volumes will continue to increase, driven by the
exploding demand and lowering price points for NAND flash technology," said
Thad Omura,
VP of marketing, Flash Components Division, LSI.
Editor's
comments:- last week I asked LSI if the improved power saving feature was
related in any way to
adaptive DSP
care. I haven't got an answer yet - so it may be the answer is No.
On the other hand maybe they're waiting for the
Flash Memory Summit (in 3 weeks
time) before they say more about their adaptive write DSP IP roadmap. |
|
LSI/SandForce have shipped over 10 million SSD
controllers - since
2010 - and
they're currently shipping over 1 million per month. | | | |
. |
Proximal Data launches
AutoCache for PCIe SSDs |
Editor:- July 23, 2012 - Proximal Data
announced
immediate availability of its first product - a
software based
SSD ASAP - designed
to work with PCIe SSDs
- and in particular those from
LSI and
Micron.
AutoCache ($999
for cache sizes less than 500GB) reduces
bottlenecks
in virtualized servers to increase VM density,
efficiency and
performance. The company says it can increase VM density upto 3x with
absolutely no impact on IT operations.
"LSI and Proximal Data have combined their respective solutions
to provide accelerated enterprise storage performance in a virtualized
environment," said LSI's director of worldwide channel sales and marketing
Brent
Blanchard. "Proximal Data's AutoCache, when used with the LSI Nytro
WarpDrive PCIe flash card, delivers explosive performance and scalability by
lowering data access latencies and resolving the VM density issue that
challenges virtualized environments. We are excited to bring this combination to
our customers."
Editor's comments:- here are some
questions I asked about the new product - and the answers I got from Rich Pappas,
Proximal's VP of sales and business development.
Editor:- How long
does it take for the algorithms to reach peak efficiency?
Pappas:- It varies by workload, but typically it takes about 15
minutes for the cache to warm to reach peak efficiency.
Editor:- Is
the caching only on reads, or is it effective on writes too?
Pappas:-
AutoCache will only cache reads, but by virtue of relieving the backend
datastore from read traffic, we have actually seen overall write performance
improvements as well. This effect is also dependent on the workload.
| | |
. |
LSI announces a new
technology roadmap for SSD accelerator components |
Editor:- April 2, 2012 - LSI today
announced
details of its new
Nytro
family of SSD technologies - which integrate and join up several previously
standalone elements in its product line in a new unified marketing direction.
In
particular LSI is saying that its legacy MegaRAID controllers and software stack
can be used as reliable proven launch pads for its
SSD ASAP /
acceleration software - which is being integrated in new upcoming generations of
PCIe SSD cards (now called Nytro WarpDrives) which use
LSI/SandForce controllers.
Editor's
comments:- in a 2009
storage market forecast I said - "the high end of the
RAID controller market
is going to disappear" - and I explained why companies in that market -
like LSI had to migrate to PCIe SSDs and SSD systems array technology such as
SSD ASAPs to satisfy the emerging needs of their oem customers - which in
previous decades had been met by RAID adapters and controller chips.
What
LSI has done in the past few years is acquire or develop individual pieces of
the technology puzzle - and selling their storage systems business
Engenio 12 months ago
so that they didn't compete with their storage oem customers - was just as
important as acquiring SandForce.
I
spoke to LSI about the new Nytro technology last week. From the sales point of
view they see this as offering affordable SSD acceleration for the masses. So
you're going to see low price point fast-enough SSD ASAPs - rather than the
fastest.
Other common features in the product line are that the
products are bootable, work with legacy
SAS software and
have minimal load on the server CPU.
LSI will also work to get better
integration between the functionality of its SSD controllers and the host cards
and caching software. That should lead to better latency and reliability in the
future.
difference between LSI and FIO?
What's the single
biggest difference you may ask - between LSI and some of the other companies in
this part of the PCIe SSD ASAP market? And in particular a company like
Fusion-io?
The
technical ingredients above are very different - and I could summarize that by
saying LSI is at heart an SSD hardware company with most of its IP in chips
- whereas FIO is at heart an SSD software company which uses chips as
deliverables - but nearly all FIO's IP is in software. That's one way of
looking at it - but the clearest difference I see between LSI and FIO is where
they are in the philosophy of their thinking re the SSD market adoption
model.
All the Nytro marketing orientation materials I saw
still talked a lot about how SSDs would fit into an HDD world.
When I
questioned that - I got the impression that LSI's corporate marketing hasn't
gone much beyond that stage. LSI is still at the "SSDs help HDDs point"
whereas FIO and many other SSD makers - and this publication - and many of you
too are beyond that and know that the
future of all
enterprise storage is solid state. The tricky part is navigating safely from
here to there.
Finally - Nytro sounds like a good name for an SSD
brand - but it's not entirely original. |
| | |
. |
SSD sudden power
loss vulnerability guide |
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
and operational
reliability.
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
negligible. |
| | |
. |
| |
.. |
Seagate to
acquire LSI's flash business for $450 million |
Editor:- May 29, 2014 - Seagate today
announced
it will acquire the assets of LSI's Accelerated
Solutions Division and Flash Components Division from Avago Technologies for $450 million in
cash.
The transaction is expected to close in the 3rd quarter of calendar
year 2014, subject to regulatory approval.
Editor's comments:-
Seagate has for a long time appeared to be a bystander rather than a
cheerleader in the SSD market.
By acquiring the
SSD controller technology
which is at the heart of a significant proportion of SSD designs in the industry
- and also the 2nd largest enterprise
PCIe SSD maker (by
volume) - Seagate will come to be regarded in a different way by SSD customers,
partners and competitors.
The cost of the acquisition - seems low
compared to many past
examples in recent market history.
LSI itself paid $370 million for
SandForce in October
2011 - and you might think that in all the years of market and product
development since then - and particularly with the controller product line
having been upgraded to being an SSD drives and module business too - the
combined business should have been worth more.
But it wasn't - and
Avago - which has owned LSI's business for a few short breaths of market time -
obviously must feel it is getting a good deal.
Possible reasons for the
low valuation of LSI's SSD business (in comparison to maybe what it might have
fetched a year or so ago are:-
- LSI was latecomer to the
adaptive R/W
controller market. So that meant many SSD companies which needed this new
type of flash management technology had already made other strategic plans for
their next generation flash management roadmaps before LSI began sampling the
SF3700 in November 2013.
- There are strong competitive offerings available now in every segment of
the SSD market - and in comparison to the past when a single controller product
could effectively be competitive in a wide variety of markets - that's no longer
true.
- Software has
become an essential part of enterprise SSD offerings. Not only does it provide
essential compatibilities - but it can also add efficiencies, reliability and
performance.
LSI's enterprise software offerings were always weak
add-ons in my view - in comparison to competing products - and were tactical
sales demonstration tools rather than genuine hard core enterprise platforms.
- Rackmounts. To succeed in the traditional enterprise SSD market today -
enterprise SSD makers need to think like systems companies or become actual
systems companies. That's where
new
efficiencies come together with new architectures.
But I can't
think of a single leading SSD systems vendor today (unlike a few years ago)
who would now offer LSI's controllers as the flagship controllers in a new
high end array.
The reasons being that LSI was late to market with
adaptive R/W, late to market with
big architecture
controller design, and doesn't have a scalable software platform.) In
a way none of those valuation theories matters - but in another way they do -
because understanding LSI's weaknesses as an SSD company stretched in too many
markets takes us neatly onto - what I think could be the real synergy between
Seagate and LSI's controller business.
What's Seagate been
historically good at?
Making huge numbers of storage drives (hard drives) cheap and
affordable.
What has the SandForce controller design been
historically good at?
Getting into more design slots for value
based enterprise and embedded applications than any other SSD controller
design.
My guess is that even if Seagate disregarded any new markets -
and focused only on the high volume potential of existing
cloud infrastructure
customers and big web entities (like Google and Baidu) - who need value based
enterprise SSDs - but who are perfectly capable of designing their own software
and APIs and firmware tweaks - then Seagate could leverage the LSI SandForce
SSD roadmaps for the next several years as a business tool to establish it as
one of (several) leaders in the utility SSD segment of the cloud.
LSI
on its own wouldn't be able to make such vast quantities of drives at commodity
prices.
Seagate on its own didn't have the SSD IP.
But the
new Seagate (with SSD IP now in its core) can leverage that IP with its
production control skills to cement new price thresholds for value enterprise
SSDs.
See also:-
hostage to the
fortunes of SSD - why big companies can't afford NOT to be in the SSD
market...
Afterthoughts (Friday May 30)
PS - Some of
the factors dampening the valuation of LSI's flash business seemed so obvious
that I didn't mention them in any of the bullet points above. But it appears
that they weren't as obvious to everyone - so here they are if you're
interested.
- Competing with your own customers.
When LSI acquired SandForce -
that introduced a new dynamic for the outlook of the combined business -
because LSI-SandForce was from that point onwards competing with many of the
companies which had previously been customers of SandForce controllers.
That
triggered plans in some of the big customer (now competitor) SSD companies to
look for alternative technology roadmaps which would level the playing field in
a future when they would be competing - at a drive level - with the controller
architecture which they had helped to establish as a de facto industry
standard.
And it also created a more favorable climate for competing
SSD controller makers who had lost out before. Such as
Marvell.
- How much are SSD companies worth?
Generally - SSD companies ain't
worth as much as they used to be.
Nearly all mainstream enterprise
SSD companies (who have been shipping products for a few years or more - as
opposed to stealthy startups) have seen their valuations dip in the past year -
due to investors realizing that the idea of any one company dominating any one
big market
segment sustainably is an illusion (for now).
That's because the
barriers to entry from new vendors are still low (due to the existence of a
very sophisticated SSD IP and parts ecosystem).
This means that
startups using new technologies at low volumes can still deliver SSDs which
are significantly cheaper to make than legacy SSD technologies - even when those
legacy SSDs are being produced in much higher production volumes. (Legacy in
this context - means anything that's more than 12 to 24 months old.)
I
suspect that the industry wide downgrade of the value of SSD companies (which
we saw last year with companies like
Fusion-io,
Violin.
Stec,
OCZ etc - for various
different reasons) will change direction - when we start to see dramatic
changes in reported revenues and profitability.
How will that happen?
I
think it depends as much on creative leadership, vision and discipline in
business models as it does on technology.
Because you can still make
money if you sell the right kind of SSD to the right kind of customer for the
right kind of purpose. (Especially if they come back soon after to buy more.)
But you can still lose money by selling an SSD that's better than it needs to
be.
And you can lose money if you over invest sales and marketing
resources by selling the right kind of SSD for the right kind of application to
a customer who won't need any more of your products for another few years
afterwards. All sounds boringly obvious business stuff doesn't it?
That's why I didn't include these add on notes in my original short version of
this post.
Another question which it's inevitable to ask - however - is
this...
What will happen to the continuing supply of SandForce SSD
controllers?
It's hard to imagine that Seagate would - from its own
self interest point of view - choose to continue supplying controllers to
such a diverse range of SSD companies indefinitely.
Users of these
controllers have been on notice that the supply outlook for these essential
chips might change ever since LSI acquired
SandForce in
January 2012.
But unlike LSI - whose main business was selling chips - Seagate's
business is selling drives - and Seagate's volume takeup of these chips (for
the reasons mentioned above) will be more than enough on its own to justify the
NRE costs - without having to sell them elsewhere.
So that may
trigger a greater sense of urgency for answers about this factor - from smaller
customers who haven't made alternative plans yet.
My guess is that if
the regulators look at this transaction - that may be one of the aspects they
will focus on - because the contiunuing supply of these controllers to smaller
SSD companies in niche markets would a good thing from the viewpoint of
competitive choice. (And it wouldn't be too onerous for Seagate to maintain
the supply for a time limited period - because Seagate's biggest SSD
competitors all have their own controllers anyway.)
Later note
added - June 4, 2014
I've been talking to a key contact in LSI's
flash business to ask if there was any official corporate statement they could
make to clarify the status of this question:- what will happen to the continuing
supply of LSI SandForce SSD controllers?
But I wasn't surprised to
learn that - despite best efforts - there is nothing more they can say about
this for now.
The reason I'm not surprised is because there are
significant external entities (regulators, customers, competitors) which could
have a bearing on the final decision and if indeed it goes ahead. So it's not
just a matter of whatever LSI and Seagate would like to do. | | |
.. |
LSI is #2 in PCIe SSDs |
Editor:- January 22, 2014 - LSI said today in
its quarterly
financial report it has shipped over 100,000
PCIe SSDs since it
entered this market in April 2012.
Editor's comments:- LSI
shipped 60,000 units in the 8 months from May to December 2013 (a figure
calculated by subtracting the numbers contained in an
earlier
announcement.) | | |
.. |
Avago wants to acquire LSI |
Editor:- December 16, 2013 - LSI today
announced
that it has agreed to be acquired by Avago
Technologies Limited in an all-cash transaction valued at $6.6
billion.
Editor's comments:- I hadn't heard of Avago before.
But I had heard of Agilent Technologies - the former name of this company -
which was a spinoff from HP. Avago was a semiconductor spinoff from Agilent.
Part of Avago's rationale (they're a semiconductor company with over 1/2 their
business in wireless technology) is to get into the enterprise storage market
and become a leader in this market "overnight".
A big chunk of the investment - about $1 billion is coming from
Silver Lake - a
VC company - which a
few years ago reintegrated another company in the SSD market -
SMART. The rest of
the funding is coming from bank loans. | | |
. |
LSI integrates "SSD
market on a chip" |
Editor:- November 18, 2013 - LSI today
launched
its 3rd generation SandForce SSD controller family - the SF3700 which - based
around a single chip design - spans a wide spectrum of SSD market
applications (from consumer to enterprise) - includes native jumper-selectable
SATA or gen 2 PCIe interfaces - and incorporates adaptive R/W DSP ECC
management.
Editor's comments:- The SF3700 (now sampling) is
the most ambitious design of a single chip
SSD controller in
SSD market
history.
Its 14 core design integrates many impressive design and
architectural features including:-
- the ability to efficiently configure as either
a small
architecture or big architecture SSD controller.
The SF3700
design can be configured with as little as 3 flash chips in entry level
consumer SSDs - or as many as 129 chips when maximally configured in a 9 channel
enterprise design which can recover from the complete failure of a memory chip
as well as partial failures in other memory chips in the array.
- dynamically adjusted power islands within the chip - enable a single
silicon design to support both the low requirements of deep sleep mode in SATA
notebooks as well as the performance requirements of entry level PCIe SSDs.
I
recently spoke to Kent
Smith at LSI about this new product.
Our conversations about
SandForce SSD controllers go back more than 4 years - so we skipped a lot of
stuff.
One of the first things I said to Kent - was - I've been
nagging you for years and asking - when are you going to do a native PCIe SSD
controller?- and for nearly 2 years it's been clear that another big hole in
LSI's SSD IP bag has been
adaptive R/W
- and now you've finally done both at the same time in a single product.
I
was also really impressed by the quality of LSI's
briefing
document on the LSI SandForce SF3700 (pdf) - which explains just about
everything you need to know. So I asked Kent - why does he need to waste time
talking to editor's like me? - why doesn't LSI just publish the document on the
web and let it speak for itself?
I said a lot of publications will
simply copy some of your pictures without attribution - and I think readers
would find it valuable seeing them too - but I think it would be fairer to the
work you've done if I could just make the whole document available - so there
was no doubt who had done the hard work of communicating what the design was all
about.
He agreed to that - and you can click on the link above to see
the original info which I got from LSI.
Some other things I learned
from this conversations were:-
- The SF3700 is a completely new design. - It leverages all the flash
related design concepts related to endurance and array level fault management
which have been proven in earlier designs and extends them too.
For
example RAISE has been enhanced so that for high-end configurations it can
protect against a full memory chip failure as well as multiple block faults -
whereas entry level SSDs which need some RAID like features but can't afford an
extra memory chip can use fractional RAISE.
- One of the reference designs which LSI offers for this controller is for an
M.2 form factor - which is goiing to be the game changing SSD for the
consumer market
next year. The card design is the same whether the SSD is being used as a
SATA or PCIe SSD. A single jumper sets the configuration at assembly time.
LSI's 2.5" reference design will also make it easier for oems to
produce products for enterprise arrays in the
2.5" PCIe SSD market. Overall
I think the SF3700 is a very ambitious and outstanding SSD controller design -
which will elevate LSI's reputation within the SSD industry.
For the
past year or so I had been wondering if the glory days of LSI's SSD controller
technology lay mostly in the past. But I can now understand why it took them
so long to integrate this new design - which is almost at the integration level
of "SSD market in a chip".
A design which integrates so many
architectural features which are optimized for so many markets wouldn't have
been feasible for a small SSD start up.
Below you can see one of the
many pictures I spoke about in LSI's paper. If you click on it you'll see the
whole thing. |
|
PS - I almost forgot to mention one funny
marketing thing I learned.
Kent
told me that they used to call the SandForce products "SSD processors"
but then found that didn't show up too well in web searches - because people
were looking for "SSD controllers" instead. So LSI has changed its
parlance and is now calling them "SSD controllers" too.
As
I said above - Kent draws some great pictures which illustrate the
functional blocks within SSDs.
He also writes a lot of SSD
blogs too. So I was relieved to hear that he gets paid on an SSD
marketer's pay grade rather than that of a writer. That means SSD editors
and bloggers don't need to worry that he'll be tempted to come and replace us. | | |
..... |
LSI says it pays to get a
2nd opinion from LDPC |
Editor:- August 13, 2013 - in a presentation
today at at the Flash Memory
Summit -
the
Nibbles and Bits of SSD Data Integrity (pdf) - LSI explained why
reserving the use of LDPC to deal mostly with read error retries (and also
later in the operating life of flash cells) can be a pragmatic design choice.
And
instead of applying different strengths of
ECC for fixed
physical block sizes - the company says another approach is to have variable
sized virtual blocks - which effectively means that better cells carry lower ECC
overhead.
See also:-
my SSD cell care scheme is
better than yours | | |
.. |
|
..... |
|
.. |
How big was the
thinking in this SSD's design? |
Does size really does matter in SSD
design?
By that I mean how big was the mental map? - not how many
inches wide is the SSD.
The novel and the short story both have their
place in literature and the pages look exactly the same. But you know from
experience which works best in different situations and why.
When
it comes to SSDs - Big versus Small SSD architecture - is something which was
in the designer's mind. Even if they didn't think about it that way at the time.
|
|
For designers, integrators,
end users and investors alike - understanding what follows from these simple
choices predicts a lot of important consequences. ...read the article | | | |
. |
|
|
| |