 |
or is it really lists?by Zsolt
Kerekes, editor - March 18, 2011 |
Editor:- If I
could just say one quick thing about the current state of the SSD market it
would be this.
Don't place too much reliance on published
benchmarks when selecting between different flash SSD vendors - even if the
tests have been set up correctly (which
wasn't always true.)
I
chose "PCIe SSDs" in the title - because they're very popular with
our readers - but my warning applies to all flash SSDs, with any
interface and in any form factor.
If I selected 3 competing models of
PCIe flash SSDs I
could design 3 comparative benchmarks in which each product ranked at the top,
bottom or middle of the 3 different top 3 lists.
I won't go into the
details - because this is a short article. But in outline the tests would
exploit or stress differences in
RAM cache
architecture, write
attenuation algorithms, and scalability characteristics.
These
scalability issues are:- how consistently each product performs with respect
to how many other SSDs are operating at the same time in the same host, the
exact choice of the host (processer architecture, clock speed, motherboard
design, installed RAM, OS and apps workload) and the in-use capacity profile
of the SSD during the test.
As I said in the first of my
11 SSD
predictions for 2011 - "SSD marketers will move away from quoting
discredited technical benchmarks like
IOPS and
closer towards promising actual Nx speedups for specific popular applications."
Sure
you have to ask yourself - is this SSD good enough for my application? (It
doesn't need to be better.)
It's just as important to ask yourself
- is this a vendor I feel comfortable buying from? Do they really want to sell
to customers like me?
That depends on how many similar SSDs you
plan to buy, over what length of time, your technical resources, support issues
etc. And how well you fit their ideal customer profile. A mismatch could lead to
a strained relationship - and after a few years - divorce. |
|
related articles
In
addition to the popular articles described in the right hand column, these
articles below also include useful sanity checks related to performance.
- SSD education
- just about every difficult topic in the SSD market revolves back around to the
the level of SSD education. Being versed in the dark arts of SSD technology
and architecture is the customer's surest defense against being taken for a ride
by vendors (who all too often don't know as much they should themselves.)
- the SSD Heresies
- is a collection of articles which will open your eyes to the many genuinely
held differences of opinion within the SSD market itself about the best way to
design SSDs and the best places to put them. After you realize that the heresies
exist it should make you feel better than you're not going nuts trying to
renconcile different facts which appear to contradict each other.
- Tuning SANs with
SSDs - although this is a new topic to many people today - this article was
published nearly 10 years ago.
- what's unique about
FIO's ioDrives? - Which of these options do you prefer? - Speeding up the
storage? - or - Speeding up the app? Find out why the raw numbers - without the
narrative - fail to tell the full story.
- RAM Cache
Ratios in flash SSDs - you won't be able to understand why an SSD which
stars in some tests is more like a dog in others. There's a simple reason.
- the New
Business Case for SSD ASAPs - auto accelerating / caching / tiering SSDs
are here for the duration - not just as a short term fix to accelerate legacy
hard disk arrays.
- RAM SSDs
versus Flash SSDs - which is Best? - far from being terminated by
flash - some RAM SSD vendors are now selling more RAM based storage systems than
ever. In some cases RAM SSDs cost less to own and run due to high floor level
entry capacities in flash and the expensive overheads incurred by making flash
storage behave like RAM storage. (And the high attrition rate of endurance.)
| |
| ... |
|
|
|
| |
... |
| For users - picking an SSD
benchmark is like picking the next card in a poker game - where you didn't get
to see the pack opened - or the cards shuffled. |
Most SSD chip companies don't
have good enough apps models against which to test their design assumptions.
And most SSD systems companies aren't technically nimble enough to
experiment with enough new radical SSD architectures in timescales which are
short enough to overlap new chip business generations.
That's why users in the market have been playing unwitting part in
Darwinian what-if experiments in SSD architecture.
I believe there's no single architecture which will suit all markets.
And I'm not alone in that view. Many SSD thought leaders don't agree on a
single unified vision for the future of solid state storage. You can read
more about these genuine fundamentalist SSD differences in -
the SSD heresies. | | |
|
| . |
|
|
| . |
|
|
| . |
|
|
| ... |
| the Problem with
Write IOPS in flash SSDs |
the "play it again Sam"
syndrome
Flash SSD "random write IOPS" are now similar
to "read IOPS" in many of the
fastest SSDs.
So
why are they such a poor predictor of application performance?
And
why are users still buying
RAM SSDs which cost an
order of magnitude more than SLC? (let alone
MLC) - even
when the IOPS specs look similar. |
 |
This article tells you
why the specs got faster - but the applications didn't. And why competing SSDs
with apparently identical benchmark results can perform completely
differently. ...read
the article | | | |
| ... |
 |
Can You
Trust Flash SSD Specs & Benchmarks? |
| Sadly no! - Many
published benchmarks for flash SSD are about as reliable as bank
valuations of Collateralized Loan Obligations (just before the onset of the
Credit Crunch). |
There are many
intrinsic technical reasons why you can't believe most published benchmarks
for flash SSDs (whether done by magazines or vendors) and why
even the tests you carefully do yourself don't give reliable results
which correlate with how the SSD will perform in real-life
applications.
We warned you of it this problem here
on StorageSearch.com - and now other publications and vendors are starting
to take it seriously too. ...read
the article | | |
| |
