will SSDs end bottlenecks? |
by Zsolt Kerekes,
editor - February 2012
Because I've been involved with the
subject of enterprise SSDs so long I sometimes find it hard to appreciate the
problems that newcomers experience when they first encounter the topic.
Judging by the emails I get - SSD newcomers sometimes conjure up imaginary
problems (and solutions) which aren't really there - or (just as often) they
don't see obvious (to me) pitfalls and dead-ends - because the outcomes
predicted by the product descriptions are cloaked in
jargon which only
experts aapreciate. (And
every word and
every concept matters and factors into the product cost and usefulness -
even in the simplest SSDs.)
Now to be fair - the enterprise SSD market
is the most complicated and confusing subject in the computer market - and even
many so called "storage experts" fail to see clearly what's going on -
because SSDs are about a lot more than just storage.
In a recent dialog
with a reader who was asking about a particular rackmount SSD supplier - I
introduced my concept of latency and position based
enterprise SSD market
silos.
There will be
7 distinct product roles
for SSDs in the all solid state datacenter of the future. (Segmented by
latency, electrical power, distance from the applications server and cost of
virtual capacity.)
Vendors - who by design (or accident) - offer
products which match the spec and price contours of those silos - will have
advantages over competitors whose products do too much, too little or don't
match the requirements in other dimensions (like physical size and price).
For
a top level view of these segments refer to the summary on the right on of this
page.
I've been talking to SSD company founders and investors about my
enterprise SSD silo classifications for about 2 years - having introduced the
concept in my petabyte
SSD roadmap article.
But unless you spend all your time thinking about the future of the SSD market -
and new SSD designs - you won't understand 80% of what we talk about.
I
know that whatever I say or write about this roadmap topic - many of the
concepts that are clear to me - will make as much sense to most people as
explanations of string theory and Higgs boson.
Analogies can't help
guide you through complex decision making either.
All they can do is
alert you to the possibility that something which looks deceptively simple from
the outside may be more complicated and have serious repercussions if you make
the wrong vendor choice.
So
- having given up on referring to architecture and market models - here's what I
said earlier today about the SSD silo concept to a reader. Some of you may
find it helpful too.
will SSDs end bottlenecks? - and cure all my
server problems
A good way to think about what SSDs will do in
the 100 per cent SSD enterprise - is to set the limits for how an enterprise
can repurpose, leverage and monetize its data - and increase process efficiency
by analyzing and anticipating customer demands in real-time.
Bottlenecks in the pure SSD datacenter will be much more serious than
in the HDD world - because responding slowly will be equivalent to transaction
failure.
And in the pure SSD storage world there will be multiple levels in the
data
driven factory of the future where SSDs which solve a bottleneck or logjam
at one level - have a cascade effect which eventually creates new bottlenecks
somewhere else - which are impsoble to fix - except by using more SSDs. That's
why SSD ASAPs
and RAM SSDs are both
permanent building blocks in the pure SSD storage architecture model - and not
just transient products (from the market life point of view) which work with
HDD arrays (ASAPs) or which have survived in some lost world which didn't feel
the impact of enterprise flash (RAM SSDs).
Having the wrong SSDs in the
wrong places will be like having a skyscraper with insufficient escalators to
move the workforce in and out each business day.
If the workers in the corporate tower can't get in and out of their
offices in a reasonable time - they might as well stay at home.
If the
right data can't get to the right place exactly when it's needed it might as
well not exist. And the enterprise which depends on that data will cease to be
viable.
And - finally - although it seems obvious to say this - if the
bottlenecks in your server system weren't originally caused by the throughput,
latency and IOPS of your storage system - but maybe by some badly written
software or insufficient CPU power or network bandwidth - or some other limiting
factor in the apps architecture - then the SSDs won't solve the problems - they
may just make them more visible.
related articles
after AFAs -
what's next? (2017)
Tuning SANs with SSDs
(2002)
auto caching /
tiering hybrid arrays
how fast can your SSD
run backwards?
Are you ready to
rethink enterprise RAM?
popular SSD articles
on StorageSearch.com
Capacitor
hold up times in 2.5" military SSDs
is remanence in
persistent memory a new security risk?
where are we
heading with memory intensive systems?
what were
the big SSD memory architecture ideas in 2016? | |
| . |
 |
| . |
|
| |
.... |
 |
| .. |
|
|
| .... |
 |
| . |
| When you start to realize
what really happens to data in enterprise memory systems the surprise isn't
that DRAM is so good. The surprise is that DRAM is so bad and it has been
getting worse rather than better for a very long time. |
| latency loving
reasons for fading out DRAM | | |
. |
| what's in a
number? |
 Editor:-
March 4, 2014 -
SSDserver rank
is a latency based configuration metric - proposed as a new standard by StorageSearch.com - which can
tersely classify any enterprise server - as seen from an SSD software
perspective - by a single lean number rating from 0 to 7. ...read the article | | |
| . |
|
. |
|
|
. |
 |
. |
DDR-4 and HMC (Hybrid
Memory Cube) are merely evolutionary increments in interface and packaging
design and don't change data architecture.
But what if you could make
RAM itself more intelligent and application aware? |
| Are you ready to
rethink enterprise RAM? | | |
. |
|
| |
