 |
| leading the way to the
new storage frontier | |
... |
| ..... |
|
top SSD companies
industrial
SSDs - boring right?
after AFAs -
what's the next box?
3D nand fab
yield - the nth layer tax?
how fast can your SSD
run backwards?
capacitor
hold up times in 2.5" military SSDs
where are we
heading with memory intensive systems?
optimizing
CPUs for use in SSDs in the Post Modernist Era
miscellaneous
consequences of the 2017 memory shortages
|
|
| ... |
|
| Nimbus
talks about SAS SSD array sauce |
Editor:- August 30, 2017 - I had another
conversation last week with Nimbus CEO Thomas Isakovich.
My curiosity was roused by the company's recent entry into the SAS SSD market.
And we had exchanged some emails about that but we both thought it would be
nice to have a longer free flowing conversation. Here are some of the things
we talked about:-
- What's the new business plan for Nimbus? Does selling drives replace
selling systems?
Tom said Nimbus will do both. He sees the high
capacity SAS SSDs as a new market opportunity.
In the past
competitors in the SAS SSD market have focused on performance rather than
capacity and cost. Due to the predicted and actual impact of
2.5" NVMe SSDs
the SAS market no longer has a performance reason to be. However even "slower"
SAS SSDs provide useful life extensions and new market roles for makers of SAS
arrays who previously used hard drives. A lot has changed in the
SAS SSD market since
the first drives were announced
10 years ago. Nimbus describe how and why they see a
gap in the
market today (pdf).
Tom said - "One of the benefits we get
from having alternative market outlets for our ExaDrive platform (the new no
write limits SAS SSD) is that more minds take a close look at the design and
operation (these being the flash drive partners Viking and SMART Modular).
This and the higher volume of drives used will result in higher quality and more
reliable SSDs compared to if we had just continued using the drives as a captive
design in our own arrays."
- Will the new drive business cannibalize Nimbus's array business?
Tom
said no.
We were both on the same page about this agreeing on all the
points we touched.
We discussed how much more
complicated
the enterprise SSD arrays segments had gotten. Nimbus's drive customers can
access as much of the software stack as they need.
The ability for
integrators to customize the array management of their SSDs has been a growing
strategic shift in the market for several years. We've seen many different way
of doing this.
One trend has been for SSD makers to customize the
controllers for their bigger cloud scale customers. But you can get better
results more easily by customizing the data management of the SSDs and array
using software which can intervene at the array and solo drive level.
Nimbus
uses standard flash controller IP to manage the flash (currently controllers
from Hynix). Nimbus's software stack provides data management which is scalable
to many petabytes in a small rack.
- How has Nimbus been affected by the memory shortages and higher memory
prices?
Here's the gist.
The
efficiency of
storage arrays - how many chips it takes to deliver storage with a given quality
of performance and usefulness - has always been important in array designs.
For as long as memory prices were falling then suboptimal
architectures could still reach customers and satisfy the business plans of the
companies which sold them. (The inefficiency hits were price, profit margin.
electrical power and rack space. But the new systems would still look good
compared to those they replaced.)
Now with constrained supplies of
memory everywhere - those designs which can do more with less - have a
competitive advantage.
Array-aware SSD controller software, coupled
with scalable fault tolerant array architecture, are the how the efficiency
problems are best managed.
For customers of Nimbus's new SAS SSDs - the
ability to use the efficiency mechanisms which come from its array level
experience will mean that its customers could do more with less chips in their
boxes.
So the shortages will be good for the market because they will
force customers to gravutate towards better designs and better software
architectures.
- Is Nimbus still averse to VC funding?
Tom said it wasn't
exactly like that because the company did have investors. The company has
followed a cautious route of aiming to be funded by its customers rather than
suck in huge amounts of VC cash which impose their own timescale pressures.
Nimbus
owns its software stack and has been cutomizing the SSDs it uses for many
years.
In effect - the multi year development and evolutionary
improvement of the company's array technologies are things you can do if you
don't have a VC checking every part of the business against a calendar which has
an IPO question hanging over every future quarter. Editor's comments:-
I've been talking to Tom Isakovich about design and architecture in storage
arrays since 2001. So we had many previous threads to pick up from which I
haven't written about here.
Nimbus - which was already a leader in the
petabyte scale SSD
market now has a viable stakeholding in the commodity SAS SSD rack market too.
The company is well placed at the intersection of several strategic pathlines to
the future
within which it can adapt and comfortably nest.
See also:-
sauce for the
SSD box gander | | |
|
| ... |
|
|
|
|
| ... |
|
 |
|
| ... |
|
| "One of the most
intriguing revelations during the Flash Memory Summit 2 weeks ago was Samsungs
new approach to stairstep etch in 3D NAND. This was one of numerous innovations
the companys EVP of Flash Products & Technologies, Kye Hyun (KH) Kyung,
shared during Samsung's eynote presentation." |
| Jim
Handy - founder Objective
Analysis in his blog
How
Samsung Will Improve 3D NAND Costs (August 23, 2017) | | |
|
| ... |
|
| SSD fires
both metaphorical and PHY |
Editor:- August 11, 2017 - From the metaphorical
point of view "fire" is generally seen as a good thing. Hot
and cleansing of dusty old ideas.
So we get
Fire
in the Valley: the Making of the Personal Computer (the book) whence I
guess arose the branding statement...
"Apple ignited the
personal computer revolution in the 1970s with the Apple II and reinvented the
personal computer in the 1980s with the Macintosh"
...a company
boilerplate (rather than a mission statement) used by the aforementioned
company in many press releases
such as this
one in 2000.
And for a while in
storage history
there was a hot USB
surrogate consumer storage interface called
Firewire.
The
hot market igniting properties of fire were an idea which - I confess - I
used / stole / borrowed to mock up a temporary (only seen by several hundred
thousand SSD readers) placeholder
banner ad for
Fusion-io in 2008
until they found the time to design some better ads. My quickly scribbled
skunkworks design used this wording... "learn more about the PCIe SSD
company company which has fired up the server market with its ioDrive
accelerators".
And "fire" has also been used as a
suffix in some SSD
product brands too:- SolidFire, WildFire etc.
But this week there
was a real physical fire at the Flash
Memory Summit.
For more about that take a look at the recent
article -
Booth
Fire Shuts Down Flash Memory Summit Show Floor on
Toms Hardware which on August 8th
said this.
"We expected some hot hardware at the Flash Memory
Summit (FMS) in Santa Clara, CA, but not like this: A booth actually caught fire
this morning." ...read
the article | | |
|
... |
|
| whatever did really happen
to ULLtraDIMM? |
Editor:- August 1, 2017 - The recent history
and market adoption of NVDIMMs is similar to the early
messy history
of the SSD market in that fascinating products appear at one time and then
fail to get traction to remain in the market in successive memory
generations.The reasons are similar:-
- Competition from other ways of getting similar work done.
In the
case of NVDIMMs not just other types of raw native memory but SSDs in other form
factors too. Such as PCIe
which can be deployed to give approachingly similar performance.
- Software support
which is meaningful.
Without a competitive and capacble software base
which can recognize the latent strengths of the new memory technology - the
results you get are never as good as the raw technology can deliver. Or if the
early software is good enough but the capability is single sourced that
deters market growth due to fears of being locked into a proprietary
supplier. A recent blog -
an NVDIMM Primer
(Part 1 of 2) (July 25, 2017) by Jim Handy - founder Objective Analysis offers
this explanation for market demise of the ULLtraDIMM.
Jim says "Both
SanDisk and IBM later abandoned the technology, which I have heard was due to
performance issues stemming from the lack of an interrupt pin on the DDR3 bus."
...read the
article
In my own contemporaneous coverage of that product I
wrote about other factors which I thought at the time indicated weaknesses in
that first generation (of its kind) product. These were:-
- there was a SATA bridge inside the DIMMs between the flash and the DDR3
logic. The result was system level performance which was not as great as you
might expect - compared to native enterprise PCIe SSDs.
- for about a year there were legal wrangles surrounding patents associated
with the design which scared off other wouldbe adopters and at one stage a court
order which stopped shipments.
- the ULLtraDIMM guzzled power - so you couldn't just drop it into a standard
motherboard socket without checking that the power tracks had sufficient
copper.
- the ULLtraDIMM product was not the "reason to buy the company"
product line in 2 successive company acquisitions of its flash technology
parents SMART then SanDisk - so it was just one of several SSD products lines
which were let go.
As you can see there were certainly enough bullets
to wound (if not kill) the first generation
ULLtraDIMM.
But its unsung (less sung about) creator had learned the lessons and produced a
superior follow up product.
related blogs on StorageSearch.com
Memory Channel
SSDs
after AFAs -
what's the next box?
NV DIMMs - the flash
backed DRAM kind
the road to DIMM
wars and Diablo's Memory1
where are we
heading with memory intensive systems and software?
controllernomics
and risk reward ratios with big memory "flash as RAM" | | |
|
... |
|
|
| |
|
|
| ... |
|
|
| ... |
 |
| ... |
Western Digital buys Tegile as memory
shortages shatter utility pricing model assumptions for hybrid storage arrays
Editor:-
August 29, 2017 - Western
Digital
announced
today that it has agreed to acquire Tegile. The price was
not disclosed.
Editor's comments:- Tegile was already using a
customer of WDC drives (SSDs, HDDs and the InfiniFlash white box SSD array from
SanDisk). So Tegile's flexible pricing models for buying storage were already a
good showcase for how to integrate these technologies in a user friendly way.
One
of the business risks of Tegile's business model was that its ideal customers
were buying usable storage based on
utility
model pricing and cost expectations which seemed predictable and scalable
upto about the tail end of 2016. Unfortunately those cost predictions have
been shattered and ruined by the rising prices of
memory
and associated shortages in 2017.
A few years ago I discussed the
risks of the utility model with Tegile - which at that time seemed to be
containable technically (because they were obviously related to expectations of
reliability, learning from similar customers doing similar things and
efficiencies). Tegile's business model meant that external finance could depend
on a predictable curve of customer value and cost.
When memory prices
rise by 50% or more (instead of going down) those curves mean that repeat
customer sales can't follow smoothly from what happened before.
Who's
to blame for the costs? Well the SSD companies which were getting most of the
benefit. So that's why I think Tegile couldn't sustain itself as an independent
company.
For WDC you could interpret this acquisition as a long
delayed response similar in thinking to
Seagate's acquisition
of Dot Hill Systems.
Tegile
can provide Western Digital with a workable platform and channel to get SSDs and
HDDs into the low end enterprise.
Analyzing where the costs should
fairly be allocated between different business units and comparing those to what
the market will tolerate while satisfying ant-trust regulators will fuel some
interesting questions for the new owner when it takes hold next week.
How
much did WD pay for Tegile?
Later:- August 30, 2017 -
WD
Expands its Graveyard with Tegile Acquisition - a blog written by AFA
competitor Nimbus - says
this...
"It's widely known that Tegile was running out of cash. No
acquirers showed up as the product is ill-suited for the cloud. Existing VC's
saw Tintri's horrific IPO and concluded that unprofitable storage companies are
DOA in the public markets. The Tegile deal was a loss for the VC's that pumped
$175 million into it (WD being among them). Rather than take the full loss, WD
paid well under $100 million to acquire it."
Netlist readies itself for next phase of SCM DIMM wars
Editor:-
August 22, 2017 - It's been an eventful past 30 days for Netlist with a
judicial
reconsideration of the merits of their memory bus load patent
more
inf (pdf),
demonstration
of their Hybridimm (with the 1st generation said to be configurable with
256-512GB NAND + 8-16GB DRAM) and
pricing
of its IPO to raise $4.5 million.
See also:-
SCM
DIMM wars news and stories
re no-show by Diablo at FMS
Editor:- August 14, 2017
- An article -
Diablo
Technologies sheds 3 execs - by The Register - lists the
recent departures of noteworthy executives at
the company and
suggests that these may be linked with the non appearance of Diablo at the
recent Flash Memory Summit.
among the awards at FMS
Editor:- August 11, 2017 -
With so many things going on in the SSD and memoryfication markets the
best
of show award winners category at the annual Flash Memory Summit has - in
past years - provided a useful way to filter interesting developments. And
this year is no exception. Among the many awards - 2 things caught my
attention:-
- Another new SSD
controller company recently
emerged
from stealth:- Burlywood.
In
the current state
of the memory market the availability of usable leading edge flash memories
is subject to great uncertainty due to technology difficulties and market
demand. And we know that flash memory makers are jealous about who they cozy up
to and share their precious early samples with.
In that context
Burlywood's claim to fame is that their IP - "Allows for multi-sourcing
of 3D TLC/QLC flash for a single SSD controller."
If you're
developing SSDs for new markets but not sure whether the memory will turn up in
your factory at the prices and volumes you require - this multi-sourcing idea
might sound attractive.
- IO Determinism - won an award for Toshiba and Facebook.
Although
by no means a new idea - because the variability and consistency of benchmark
related factors such as latency in enterprise SSDs has been discussed in
these pages and
elsewhere ever since
flash SSDs have been used in the enterprise - nevertheless the recent award
has refreshed the idea. And instead of the
competitor A versus
competitor B arguments which were popular in the PCIe SSD market 6 years ago
- the new comparison at the heart of the award showed the difference in the
quality of latency in NVMe SSDs from the same company - due to firmware. read
more about it You might say that what the 2 different awards
above share in common is the desire for predictability in environments
which are beset by highly chaotic elements.
Nimbus enters the SSD controller market
Editor:-
August 10, 2017 - Nimbus
Data Systems recently
announced
it has entered the SSD
controller market with a reference design for high capacity
SAS SSDs.
Editor's
comments:- this move is part of a strategic trend in the market. For more see my
new blog - sauce
for the SSD box gander
Gen-Z does memory fabric demo at FMS
Editor:- August
8, 2017 -
Members
of a relatively new ORG -
the Gen-Z Consortium ran
multi-vendor demonstrations this week at
FMS achieving 112GT/s.
According
to Gen-Z's faqs page
- the idea is to create a high bandwidth, low latency, standard for
memory-like data transfers which are media independent and can "scale
from tens to several hundred GB/s of bandwidth with sub-100ns load-to-use memory
latency."
What's a comparable context?
If you go back
in time to 2000 and think about the past but forward looking potential of
Infiniband or back
in time to about 2013 with
PCIe fabrics
- it's maybe a bit like like those were in their time - but now we're looking
from a 2017 competitive needs analysis and the memoryfication of the
datacenter- so it needs to be faster.
Intel promises dual port SSDs
Editor:- August 7,
2017 - Intel
today
announced
some new upcoming data center oriented SSDs among which were:-
- dual port versions of some SSD models promised in Q3 2017
- a new form factor for SSDs - the "ruler" form factor - which the
company said "will come to market in the near future".
See
also:-
"dual
port SSD" - mentions on the mouse site
Longsys showcases tiny NVMe BGA SSDs at FMS
Editor:-
August 7, 2017 - Longsys
today
announced
that it will be sampling the industry's first 11.5x13mm NVMe BGA SSDs which
support Boot Partitions and the Host Memory Buffer features of NVMe rev 1.3
in Q4 2017.
The new
P900 series -
skinny
(DRAM-less) PCIe 3.0x2 SSDs - aimed at the
consumer market
and being shown in the next few days at the
Flash Memory Summit - will be
available in capacities from 60GB to 480GB (64-layer 3D NAND ) and use the
88NV1160 controller from Marvell.
"As the world's leading Mini SDP SATA DRAM-less SSD module
provider, Longsys has once again made a unique contribution to the storage
industry by pioneering the integration of Marvell's DRAM-less SSD controller
88NV1160 and the new 64-layer 3D NAND into a 11.5mmX13mm BGA package," said
Zhixiong Li, CTO of Longsys.
Editor's comments:- apart from
the obvious applications for NVMe SSDs on a chip in the consumer market I think
the widespread availability of such devices in 2018 could create new
opportunities for disruptive "high availability" BGA array
architectures in industrial market too.
See also:-
SSDs on a chip,
PCIe SSDs
SMART sets no write limits on new 50TB SAS SSD
Editor:-
August 4, 2017 - SMART
Modular Technologies today
announced
that it will demonstrate a new 50TB dual port SAS SSD in 3.5" form factor
next week at FMS.
SMART
says that its new
Osmium Drive
consumes as little as 1/8th the power per terabyte compared to nearline HDDs.
Noteworthy from the
endurance
and DWPD point of view
SMART's
datasheet
(pdf) says the MLC drive is specified to offer " unlimited writes"
over a 5-year lifetime. Behind this may be the technical judgement that at SAS-2
speed and this storage capacity you can never write too much data.
Editor's
comments:- in April
2013 - a related company - SMART Storage Systems was offering a 2TB 2.5"
SAS SSD rated at 10 DWPD as a capacity leading fast storage drive for under
$4,000. That company and product line was acquired by SanDisk. The SMART Modular
Technologies in today's story is the industrial and special memory SSD business
units which remained unaquired by SanDisk. So it's interesting to see them
going back into a market where they effectively compete.
SAS is no
longer regarded as a high performance drive interface (that role has long
been taken over by PCIe and memory channel channel devices). Instead SAS has
changed its role into being a convenient internal technology for high capacity
fault tolerant rackmount SSDs which deliver
multipetabyte
scale SSD storage.
For other competing companies which have also
announced ultra high capacity SAS SSDs see the
SAS SSD news archive.
Mobiveil releases FPGA controller for 16 lane 16G NVMe SSDs
Editor:-
August 3, 2017 -
Mobiveil today
announced
availability of its FPGA-based SSD development platform targeting the latest
3D NAND devices. Error correction is performed using ether BCH or LDPC.
The
NVMe (rev 1.3 compatible) controller supports a multi-port configuration for
efficient I/O virtualization and multi-path I/O and namespace sharing.
Mobiveil's silicon-proven PCIe solution has added Gen4 support for up
to 16 lanes at 16G line rate with availability of 512 bit Data path user
interface. The PCIe controller offers AXI4 interface and DMA capabilities for
seamless integration into an ARM Advanced Microcontroller Bus Architecture SoC
implementation.
Micron Awarded Virginia Values Veterans Certification
Editor:-
August 3, 2017 - Micron
today announced
that its site in Virginia has been recognized for its services which support
veterans within the company's hiring program. Among other things Micron said...
"More
than 10% of Micron Technology Virginia site's nearly 1,300 full time team
members are military veterans. Micron has also funded the creation of two
internal military service employee resource groups, providing veterans with peer
support opportunities to help diminish some of the difficulties that
transitioning to civilian life can bring. At the corporate level, Micron
connects with local military bases across the country through TAP (Transition
Assistance Programs) and actively participates in the Hiring our Heroes program."
See
also:-
why
was it so hard to compile a simple list of military SSD companies?
Lessons from Coraid
Editor:- August 2, 2017 -
AoE (ATA-over-Ethernet) is
one of those ideas from
storage history
(AoE news coverage here on the mouse site started in 2003) which flash into
the news pages for a while and then fade away.
There's a new
article
on StorageNewsletter - interview with Brantley Coile, CEO and Founder, Coraid
- which takes you into the thinking behind the original concept and brings you
up to date with what his new company is doing now.
Among other things
in the article - Brantley
Coile talks about the mistaken business direction which Coraid took to
grow revenue after its successes in 2010 to 2012...
"...then
management made a huge mistake: they decided to change the operating systems.
Instead of using the Bell Labs technology I had used, and still use, they wanted
to switch to Solaris. Companies like Coraid can't afford to change OSes no
matter what the reason. It confuses customers. It means completely changing
developers. It stops new features as the new team relearns all the lessons the
old team had already discovered. Sales sagged. Funding disappeared." ...read
the article
See also:-
VCs in storage ,
Surviving the
Solaris x86 Wars
AccelStor's new 100GbE HA NVMe rackmount SSD is 6x faster
Editor:-
August 2, 2017 - AccelStor
today
announced
a new high
availability all-NVMe flash storage array, the NeoSapphire H810. The 4U
rackmount SSD which uses Intel's Xeon Purley platform and has 100GbE
connectivity delivers upto 6x the performance of the company's previous
NeoSapphire P310 all-NVMe flash array will be on display next week at the Flash Memory Summit.
See
also:- the fastest SSDs,
high availability
SSDs
more funding for Primary Data
Editor:- August 2,
2017 - Primary
Data today
announced
it has raised a total of $40 million in its 2nd round of funding to support
continued sales growth in tandem with the 2.0 release of its DataSphere software
platform. | |
| . |
|
|
| . |
|
| . |
|
| |
| .. |
|
|
.. |
|
|
|
.. |
|
| SSD news in
Augusts of yore |
- August 2007
- STEC announced it
would ship the industry's first native SAS flash SSDs in Q1 2008.
- August 2012
- Skyera launched
a 1U half-depth 10GbE SSD rack with 44TB usable capacity ($131,000 approx),
3.6GB/s throughput and upto 1 million IOPS using under 800W electrical power.
- August 2015
- Everspin announced
the world's first all MRAM storage module in the M.2 form factor.
- August 2016
- Plexistor showed
it could handle millions of remote writes per second using its persistent
memory over fabric software.
| | |
|
.. |
|
"The unexpectedly
higher price of DRAM and nand flash in the past several quarters due to demand
and yield issues has been like manna from heaven to companies with alternative
nvms.
The change in relativistic competitive landscape has had the
same effect as if the alternative nvms could time travel 2 years into the future
while nand and DRAM have stayed looped in Groundhog Day." |
| Zsolt Kerekes,
editor - StorageSearch.com
commenting on -
Making money
is so DRAM easy for some memory-flingers - the Register (August 17, 2017) | | |
|
.. |
|
The main performance
benefit of HMB (use of host memory buffer) in RAMless consumer NVMe PCIe SSDs is
random reads (40 to 70%) and burst writes (upto 5x).
But multi-level
eror checks and protection are needed in the controller to ensure data integrity
in the flash SSD in the event of a surprise disconnection from host memory. |
| Notes from the paper -
Improving
the Design of DRAM-Less PCIe SSD (pdf) - by Sean Yang,
Product Manager - Phison. at Flash
Memory Summit (August 2017). | | |
|
.. |
|
|
|
.. |
|
|
|
.. |
|
|
|
.. |
|
| after AFAs?
- the next box |
Throughout the
history of
the data storage market we've always expected the capacity of enterprise user
memory systems to be much smaller than the capacity of all the other attached
storage in the same data processing environment.
 A
new blog on StorageSearch.com
- cloud
adapted memory systems - asks (among other things) if this will always be
true.
Like many of you - I've been thinking a lot about the
evolution of memory technologies and data architectures in the past year. I
wasn't sure when would be the best time to share my thoughts about this one.
But the timing seems right now. ...read the
article | | |
|
.. |
|
|
|
.
. |
|
|
|
| Are we
there yet? |
| After more than 20 years of writing guides to
the SSD and memory systems market I admit in a new blog on
StorageSearch.com -
Are
we there yet? - that when I come to think about it candidly the SSD
industry and my publishing output are both still very much "under
construction". ...read
the article | | |
|
.
.
.
. |
|
 |
|
.
.
.
. |
|
| RAM has changed from being
tied to a physical component to being a virtualized systems software idea - and
the concept of RAM even stretches to a multi-cabinet memory fabric. |
| what's
RAM really? - RAM in an SSD context | | |
|
.
.
.
. |
|
| If you're trying to predict
and anticipate how the supply of next generation nand flash will ramp up in the
next year compared to how you've seen memory successions before then the 3D nand
flash market has presented many problems of analysis and interpretation. |
| 3D nand successions?-
more dimensions of analysis | | |
|
.
.
.
. |
|
|
|
.
.
.
. |
|
| I think it's not too strong
to say that the enterprise PCIe SSD market (as we once knew it) has exploded and
fragmented into many different directions. |
| what's changed in enterprise
PCIe SSD? | | |
| . |
.
.
.
.VCs
in SSDs
SSD
controllers
the SSD news
archives
Can you
trust SSD market data?
what's
RAM really? - RAM in an SSD context
Can you tell me the best
way to get to SSD Street? |
|
| |
