 |
Debunking Tier 0 Storage Babble
May 18, 2009 - by
Zsolt Kerekes, editor
StorageSearch.com |
"Tier 0" is
not a term I've ever knowingly used in editorial - which you may find strange
for a storage publication which
talks
so much about SSDs.
It's not that I'm unaware of its existence. I
just resist any jargon
term which contributes no useful new concepts or (as is the case with tier 0)
is completely misleading and doesn't tell you anything you didn't already
know.
Don't get me wrong. I've got nothing against recognizing that
traditional storage hierarchies need to add new levels.
I remember
telling a bunch of bunch of HSM vendors back in 1999 that their pretty
little pyramid shaped diagrams seemed to lacking a level or 2 for SSDs.
"What's
an SSD?" I was asked. OK that was 10 years ago, and they might
have had an excuse for not knowing back then. The HSM companies didn't exactly
set the world on fire because they were mainly backwards looking. They knew
how to tell the difference between a
tape drive,
optical jukebox and
hard disk and if you
bought their software - the supplier had gone bust or acquired before you
installed version 2.0.
HSM
stopped being a sexy term in press releases round about the same time that NAS
came into vogue.
You may be surprised that as recently as a few months
ago (in 2009) I was amazed by a storage vendor who was so proud of the ability
of their software to help you manage the difference between 7,200 RPM and
15,000 RPM storage arrays. I asked what it could do about 300,000 RPM equivalent
storage - like SSDs? "It can't do those" - I was told. I won't name
the company here - because I'm not trying to embarrass anyone.
Sometimes
I'll adopt a new jargon term the very 1st time I see it. A recent example of
that was "write
amplification".
On a slow news day I'll poke fun at companies
who try to twist and chisel perfectly good words (which are understood by
nearly everyone) into newer varieties, with a slight nuance of meaning.
There's a list of examples on the
RAID systems page if
you're interested.
Green storage? -
Hmpf.
Mostly
when a new phrase comes along - I'll wait and see how many other companies
adopt it before running with it in a big way.
So why have I resisted
the term "tier 0 storage" for so many years?
What have I got
against it?
My answer comes in 2 parts....
1 - What's
wrong with "SSD"? - A simple word which is starting to become
better understood - and is in fact the #1 search term which brings readers to
this site. (OK we know that SSD is preloaded with many cultural values and
assumptions - depending on who uses the word. And we also know that not all SSDs
are created equal. That's a topic of many articles.)
2 - Tier 0 is
misleading. It gives you a simplistic and unrealistic view of the storage
acceleration concept.
Suppose you plug in a
PCIe flash SSD card to
a server. I think most tier 0 fans would agree that's a good example of what
their beloved term is about. (Other variations can also include SSDs with
other interfaces installed inside or close to the server box.) So far - so good.
The result - in a well designed system - is a server which can run some
applications faster than the same server using rotating storage. That's the
principle I call SSD CPU equivalence. It's not a new concept BTW. Just a
newish phrase for something I discovered in systems I was integrating 20 years
ago - and was known to many others - long before then.
At this
point.... We've successfully used an SSD and the term "tier 0" in a
context which seems clear and unambiguous.
Now it will get
complicated... What happens when you have 100 or so of these "tier 0"
inside servers operating in an application where they have to share or exchange
some critical data?
A new bottleneck creeps in - on the
SAN. (I'll call it the SAN
to keep it simple - but actually the storage interconnect between the 100 or so
servers could be Ethernet,
InfiniBand or
whatever.)
As we know from hundreds of published (and more unpublished)
case studies the
fix to this type of problem is to attach a faster SSD (typically a
RAM SSD) to unblock the
dataflow logjams across the storage network.
So now we have a new box
- the dataflow logjam unblocker - which may internally have a latency and
bandwidth 10x or 100x faster than the SSDs - which we earlier called "Tier
0 storage".
What are we going to call the new SAN SSD? - Derrr...
"Tier minus one?" or "Tier a bigger number?"
You
can see my quarrel with "Tier 0". It assumes a very simple
architectural model. And extrapolating from the term takes you nowhere useful
from a semantic or technical point of view. It really doesn't tell you
anything that you didn't already know before.
In real life you can
have multiple levels of SSD at the DAS level - inside the server box - and
externally too. It's going to be hard enough learning about all the many flavors
of SSD. So let's not waste any of our precious brain cells by investing "tier
0 storage" with an importance it doesn't deserve. | |
... |
|
|
| . |
|
|
| . |
| the Problem
with Write IOPS |
Editor:- December 16, 2009 -
StorageSearch.com today
published a new article -
the Problem with
Write IOPS - in flash SSDs.
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
9x more than SLC? - even when the IOPS specs look similar. |
 |
This new 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 | | |
| . |
|
|
| . | |
