click to visit home page
leading the way to the new storage frontier .....
the fastest SSDs - click to read article
the fastest SSDs ..
image shows mouse battering down door to cheese store - click for RAM SSDs  directorypage
read the article on SSD ASAPs
auto tiering SSDs ..
image shows megabyte waving the winners trophy - there are over 200 SSD oems - which ones matter? - click to read article
top SSD companies ..
SSDs over 163  current & past oems profiled
SSD news ..

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
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?
SSD ad - click for more info

storage search banner

data storage  latency and bottlenecks
SSD articles
SSD reliability
what do enterprise SSD users want?
latency - SSD bookmarks from IBM
how fast can your SSD run backwards?
High Availability enterprise SSD arrays
is it realistic to talk about memory IOPS?
characteristics which limit flash SSD performance
what do I need to know about any new rackmount SSD?
controllernomics and benchmarks in flash tiered as RAM
SSD ad - click for more info
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?
SSDserver rank article imageEditor:- March 4, 2014 - SSDserver rank is a latency based configuration metric - proposed as a new standard by - which can tersely classify any enterprise server - as seen from an SSD software perspective - by a single lean number rating from 0 to 7. the article

flash SSD capacity - the iceberg syndrome
Have you ever wondered how the amount of flash inside a flash SSD compares to the capacity shown on the invoice?

What you see isn't always what you get.
nothing surprised the penguins - click to read  the article There can be huge variations in different designs as vendors leverage invisible internal capacity to tweak key performance and reliability parameters. the article

SSD ad - click for more info

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?

There's a genuine characterization problem for the SCM industry which is:- what are the most useful metrics to judge tiered memory systems by?
is it realistic to talk about memory IOPS?