You could
easily get the impression that this is going to be a clone market - same host
interface - same memory chips -who cares who makes them? Only
price matters -
right?
That's what I expected would be the case when I started
discussing the PCIe SSD market with vendors 4 years ago.
Back then
(in
2007) I
thought the PCIe SSD market would end up being dominated by low cost vendors
based in Asia - because I underestimated how difficult it would be for designers
to manage
reliability and
performance. I
expected to see the market converging into a bunch of similar capability PCIe
SSDs - differing only in
price.
Instead
what we're seeing today are tremendous differences in design features such as:-
how many PCIe SSDs on the same host? - some designs
scale better
than others
how hard is it to integrate the PCIe SSD with the apps software? - some
products now support auto-tiering
and are optimized for use as cache - while others can be used flexibly and
viewed as either fast storage or cache (or both) depending how the system is
configured
how much of the host
CPU power is needed to make the SSDs work? - this is important if you're
trying to fix an already overloaded production server - because you can't
afford to lose performance while you tune the hot spots (even if the theoretical
end point of the tuning process is faster)
how good are the reliability
features? - if a given percentage of the raw flash fails some designs will carry
on working - others won't
what are the field upgrade and
fault replacement
logistics? - how easy is it to install or remove new modules and starte them
running?
what's the size? - PCIe SSDs come in tradional cards, multi-slot modules,
and also more recently they have been demonstrated in
3.5" and
2.5" form
factors (SCSI Express).
Today's market leaders are those with the best
SSD controller technology
and architecture rather than companies which have access to the lowest cost
chip assembly lines. And by "architecture" I don't just mean
what's done in silicon.
"Architecture"
is what was in the mind of the designer and includes more or less elegant ways
of dealing with all the application stresses that the product will encounter
during its market life such as balancing the
11 key SSD design
symmetries and choosing whether to go with
big vs small
memory array architecture.
Software and API
stability play a key part here too.
In this corner of the advancing
enterprise SSD tide - the role of solid state storage is starting to fuse into
the apps servers - and SSDs play a key role in determining what those
particular servers can do. It won't be too long before the SSD count becomes
just as important in server marketing as the clock speed of the CPU and the
number of cores.
If I try to look ahead at what these PCIe SSDs are
going to look like - I don't see a picture of memory chip arrays connected to
a processor bus through cloned controllers. Instead I see greater differences
emerging than we're seeing today.
The pictures in the ads may look the
same - but what they do - and how they do it - are differences we'll have to
learn more about.
In many ways the enterprise
SSD market resembles a jungle and the creatures within it are evolving fast.
The bugs haven't read the blurb on the bug spray - and don't realize they are
supposed to stay away - or be dead.
As you know from market
history - the proposition that one new SSD thing can replace one old SSD thing
is rarely as simple as the advocates of the new thing say.
