|
|
 |
DensBits
samples new TLC flash controller
Editor:- April 30, 2012 - DensBits today
released a new SSD controller - the
DB3610 -
which supports the latest 2Xnm and 1Xnm TLC (3 bits/cell) MLC flash with an
extreme endurance
figure of more than 10K P/E cycles and R/W performance of up to 95MB/s / 65MB/s
and 4,000 / 1,100 R/W
IOPS (4KB),
for sequential and random operations, respectively.
DB3610 employs
DensBits' Memory
Modem technology (adaptive DSP in SSD IP) which enables a native TLC
solution with more than double the endurance of 2 bits/cell (MLC), and near-MLC
R/W performance.
Editor's comments:- It's easy to miss the
significance of new SSD products and technologies. And you might think from
looking at the text and numbers above - this is a consumer style
SSD controller - and
it's not for me.
But I think DensBits may become one of the
top 20 SSD companies
real soon - unless it gets acquired before that happens. Its flash technology
has very high roadmap
symmetry and the potential to impact competitiveness in the consumer,
embedded and fast-enough enterprise SSD markets with a splash that's as big as
SandForce made
when it emerged on the scene 3 years ago. You can read more in
Who's who in SSD? -
DensBits.
SandForce driven SSDs get 5x SMARTer
Editor:- April
26, 2012 - SMART
is close to sampling a new
fast-enough (enterprise
component) 2.5"
SATA SSD - called
the CloudSpeed 500 - which uses a controller from LSI/SandForce - and
I can already see a lot of you starting to yawn or click away or look at
something different on this page or out of the window - which is a good
argument for putting
duct tape over that camera lens on your browsing device BTW.
"Let
me cut you off right here!" - I said to SMART's President - John
Scaramuzzo last week - right after we had done with the conference bridge
pings and greetings. "Because if all you want to talk to me about is
just another
SandForce
driven SSD then this is going to be a very short conversation. Can we talk
about something else instead? - because this is not going to get onto my news
page - which has a big yellow note on it explaining why."
Just to
drive home my point I followed up by saying - "There has to be something
different in a SandForce inside announcement to make it news for me and my
readers. So for example - if STEC
or SanDisk launched a
SandForce based SSD that would be news."
John said - "I saw
your note - and we have a new twist on this which no one has done before."
So
I listened. He was right. Here goes...
SMART have learned a lot -
about better ways to interact with cheap consumer flash - from the experience
they accumulated designing the
adaptive
DSP based SSD controller which goes inside their Optimus SSD (which I wrote
about in February).
That gives them raw stats for factors like
write pulse length,
endurance
and data integrity
- when they are in this closed loop adaptive DSP managed flash environment -
with brand X flash memory and generation Y process. In a new wave DSP SSD
controller those figures are going to be different in different parts of the
memory and different over time - but it gives you an idea of what the memory can
do. (With a view which even the original memory manufacturer doesn't have.)
What
SMART have done is ask - if we know this stuff about this type of memory - and
precondition it during manufacture or first boot (with different parameters
to the default setting done in the fab by the chip maker) then can we get better
reliability even when using an SSD controller which doesn't see the memory the
same way and runs open-loop?
They found the answer is - yes! They can't
get exactly the same reliability that they would get with their Optimus
controller - but they do get a 5x reliability boost by using the new
setups with unmodified and lower cost SandForce chips.
SMART has
to back off the preset numbers - to allow a safety margin - because they're
optimizing the numbers for the whole population of memories they will use -
rather than optimizing each part of each memory.
To use an analogy
here - your average safe running speed is different for you as an active kid
compared to you with a broken ankle or you at 3 score years and 10. And the
average safe running speed for an olympic athlete is different to the
population as a whole. Traditional SSD controllers can't tell if this piece of
flash is an olympic champion or not - and unlike new wave controllers - they
can't give them juice.
And here's another analogy. I changed my
behavior in this interview with SMART from what I thought I would to - to do
something different instead as a result of what I learned. (And I hope it was
worth it for you.)
SMART's trick with the SandForce controllers is like
using Dolby correction with a 1980s cassette tape. Whereas SMART's trick with
its Optimus controller is like having a dynamic sound equalizer built into an
iPod.
What's this got to do with SSDs? There's a new wave of flash SSD
management - based around combining adaptive flash interaction with DSP. In one
way it's not that new - because some companies have been working on it for
years. What's new is that more vendors are doing it - and in different ways. In
order to create enough space to discuss these issues more fully for those of who
are interested - StorageSearch.com will publish a new directory for housing
content related to adaptive DSP flash management in SSD IP in the next few
weeks.
Now coming back to the "cloud" thing.
Attaching the word "cloud" to an SSD name isn't new. I recall that
OCZ have done it before -
and there are many other examples. Maybe there should should be a new editor's
filtering rule about SSDs and cloud names. Don't look now but here comes
another one...
Diverse views emerge on SSD caching / tiering video
Editor:-
April 26, 2012 - In a 50 minutes panel discussion YouTube
is SSD simply a big cache
or a real storage tier? - moderated by Chris Evans, editor of
the Storage Architect.
Participating
in this discussion on
SSD ASAPs were
representatives from EMC,
STEC,
Nimble Storage,
Velobit,
SolidFire and
Virident.
The
discussion ranged from - where's the best place to put SSDs? and which agency
should determine where to put the hot data? The app or the storage system?
Here are just some of the ideas expressed in the talk.
Caching is best
because it's simple to deploy and scale - and you don't have to bet your
career or your business future
if it goes wrong -
(and discover new
bottlenecks
which you didn't know about before).
Caching is a short term fix. SSD
is a tier.
You need both caching and tiering.
SSDs provide new
possibilities which means maybe the SSD shouldn't be hidden by caching - but
remain visible to SSD specific APIs.
You don't see any of the big HDD
storage box makers in the biggest customer data centers like Amazon, Google and
FaceBook. The HDD isn't a datacenter device anymore - hard drives just live in
the cloud.
See also:- the
SSD Heresies
Nimbus publishes tick test results
Editor:- April
25, 2012 - Nimbus Data
Systems today
announced
that several key performance and operational characteristics of its
S-class
systems (iSCSI
rackmount SSDs)
have been validated by Demartek.
- Throughput:- a single Nimbus S-Class 2.5 TB system with a dual-port QDR
Infiniband
connection delivered (near line-rate) 7.6 GBps performance on reads and over
2GBps on parity-protected (RAID
5/6) writes.
- Support of Automatic SSD Enablement, a
new
feature in vSphere 5 (pdf) that leverages the low latency of flash
technology to improve VMware operations with simplified out-of-the-box
administration.
Editor's comments:- it's a complicated
business doing meaningful SSD
tests which can be used as an input into performance modelling. And I've
seen many vendor funded SSD test reports which failed in that respect. But
recently - as the market has got more experienced - some SSD vendors are
changing the emphasis of their sponsored reports to show that their products
can walk and chew gum at the same time. That's the message I pick up from this
Nimbus press release. How much gum? And how brisk the walking pace? It will
suit some users more than others.
Permabit launches SSD dedupe software
Editor:- April
25, 2012 - -Permabit
today
launched
a low latency
dedupe
engine (pdf) which has been
optimized
for flash SSDs and which is scalable to millions of
dedupe operations
per second.
The product is aimed at SSD appliance makers.
still waiting to learn more about GridIron's SSDs?
Editor:-
April 24, 2012 - GridIron
Systems yesterday
announced
that it is approaching
petabyte scale
FC SAN SSD capability
with its OneAppliance auto-tiering
product family which will be shipping next month.
Building blocks
include:-
- FlashCube (1 million IOPS, 10GB/s R/W, upto 100TB)
- iNode (40 core TB RAM with 100TB flash)
- RackPack (40 server system with 4 million IOPS, 40GB/s bandwidth and 250TB
of Flash
Editor's comments:- I took the
unusual
step (for me) of registering with GridIron's website so I could read the
product details. I had asked the company last December to remove this
anti-informational inquiry process - but the barrier is still in place. Today
it took far too many many minutes for their web site to come back with a
message saying that I could now be allowed to read their datasheets. But by
then I had used up my time budget.
GridIron's
blogs suggest that because their
systems are so fast - you'll get better results using their racks for
big data analytics instead of other SSDs - and in particular
PCIe SSDs like those
from Fusion-io,
OCZ and
LSI.
Are you
going to wait 5 minutes or more on their web site to get information which you
should be able to see immediately so you can decide how credible is their claim
that they can save you time and money with their proprietary solution?
Someone
should tell their VCs that this company is still in stealth mode when it comes
to web based communications.
And
SSD racks versus PCIe
SSDs is a flawed analysis anyway. Most enterprises will need and use
both.
...Later:-
April 28, 2012 - GridIron's CTO, Som Sikdar responded to my
criticism above and said the company
is
sorry and will review and improve the information accessibility on its
website. And a few days later - the log-in wall was removed.
how fast can your SSD run backwards?
Editor:- April
20, 2012 - StorageSearch.com
today published a new article which looks at the
11 key symmetries in
SSD design.
Whether you're a
buyer,
designer,
marketer or
investor in SSDs -
this new conceptual framework will help you to comparatively evaluate any SSD
compared to competitive offerings. It's helpful whether you're looking at raw
SSD IP and controller chips right up to the most complex
datacenter SSD
racks.
SSDs are complex devices and there's a lot of mysterious
behavior which isn't fully revealed by
benchmarks and
vendor's product datasheets and whitepapers. Underlying all the important
aspects of SSD behavior are asymmetries which arise from the intrinsic
technologies and architecture inside the SSD.
Which symmetries are
most important in an SSD? - That depends on your application. But knowing
that these symmetries exist, what they are, and judging how your selected SSD
compares will give you new insights into SSD
performance,
cost and
reliability.
There's
no such thing as - the perfect SSD - existing in the market today - but
the SSD symmetry list helps you to understand where any SSD in any memory
technology stands relative to the ideal. And it explains why deviations from the
ideal can matter.
The new article unifies all SSD architectures and
technologies in a simple to understand way. Now that I've spent
20 years thinking about
it - it all seems really obvious now. This is the most important article
about SSDs that I've written in the past few years. I couldn't have written
it before. I hope you like it.
...click to read
the article
PS - I'll be using the new symmetry terminology
from now on in news stories and in SSD company profiles - so you'll soon get
used to it - just as you've got used to lots of other
SSD jargon.
RamSan meets SANsymphony to negotiate SSD ASAP
Editor:-
April 18, 2012 - Texas
Memory Systems today
announced an
auto-acceleration /
SSD ASAP marketing bundle with DataCore Software.
"Companies
are looking for solutions that make SSD benefits available across the entire IT
infrastructure..." said Dan Scheel, President of
Texas Memory Systems. "Rather than go with a rip-and-replace upgrade to
get these benefits, the TMS/DataCore bundle will provide better utilisation of
existing storage infrastructure while dramatically improving it." |
|
|
| |
|
