Availability and Control
Everyone with an interest in data and data
management will know that there are scaling levels of availability and
assurance. With increasing availability levels comes a decreasing control over
exactly where your data resides and, to some, a security risk. Let’s consider a
humble word document (other document types are available) and your control over
it as it progresses up the availability tree.
Location
|
Availability
|
Control
|
Hard
drive (Laptop or PC internal/external USB etc.)
|
Back it up. Lose it and
it’s a restore from last known good backup, sometimes with less than optimal
results.
|
It’s there, in front of you
and you can literally see the external drive itself and know that your doc is
in there, but depending on where you backed it up to you could potentially
lose it in a disaster situation[1].
|
Simple
RAID
|
RAID two drives together (1
or 10), lose one and you still have another one. Also back it up.
|
You’ll still be confident
that your doc is on both those drives and under your control but, again
depending on where you back it up, you could still lose it in a disaster
situation.
|
Storage
Array
|
More complex raid levels
(5, 6, DP etc.) so you can potentially lose more than one drive and still
have data availability confidence.
|
Starting to get a little
hazy on exactly where your data is. It’s on those hard drives but accessing
those drives individually you’d really struggle to get your data back.
|
Multiple
site storage arrays
|
RAID levels locally which
will keep your doc safe and will replicate it across to another site some
distance away. Alternatively, the whole array is clustered so it is “highly”
available on both sides.
|
Seems a lot of expense and
effort for a little word document …
So exactly where is that
doc? Does it exist on both sides? How different are they, in terms of a point
in time? Can you put your finger on exactly where it is at any given time?
Doubtful.
|
Private
Cloud
|
Multiple arrays across
multiple locations that could be being served, storage wise, from one of any
number of platforms. Loss of an entire site can be tolerated.
|
You can still cling to the
fact that your doc is still in one of your datacentres somewhere but it is a very indistinct. The precise location is
all but impossible to discern from the end user point of view
|
Public
Cloud/Distributed
|
AWS, vCloud, Azure etc.
potentially utilising object oriented storage[2]
and erasure coding to keep data as available as it can possibly be,
recoverable even if you lose multiple sites
|
Your doc now exists as a
somewhat ethereal entity. If erasure coding is in play to ensure data is
available in the event of multiple site failures, your word doc is pulled
apart, combined with other hashing data and spread out over multiple
locations. Any single piece of this will be unrecognisable. Try to reassemble
it in flight and it will be unrecognisable.
|
Or …
With increasing availability your word doc
moves further away from being under your direct control. Data sovereignty considerations
aside, if you want to keep your data in a known state and location, you’ll need
to compromise on availability. There’s an intersect somewhere between the two
storage options that many organisations occupy, somewhere around the dual site,
replication, clustering sort of area. These organisations also increase control
with security measures like in-flight and
at-rest encryption technologies that
work happily alongside distributed data environments, but nonetheless you can’t
honestly say “I know exactly where my data is”.
Clearly, knowing the exact location of your
data may not be such a good thing in some situations. You can be compromised
into divulging the details of your data if you know its’ exact whereabouts but
the idea of giving up control of your data to a third party or to technology is
still anathema to some.
Control is an illusion
Control is
an illusion. No matter how much you think you can control your data there will
always be scope for something to take a situation out of your control, so how
would you lessen the impact of the relinquishing control? Let’s add another
line to the little graph – responsibility.
It follows the availability line closely
because, as you give over control, responsibility for loss of the data goes
with it. Although you won’t ever lose all responsibility even when you’ve given
over complete control, you no longer have the additional overhead of managing
the data yourself and being responsible, financially or technologically, for
its well-being. Perhaps more importantly, along with a diminished likelihood of
data loss, there is greater scope for recourse if it does happen with contracts
in place such as service level agreements, uptime guarantees and the like.
Availability is real
At the risk of everyone now having a
certain song in their head for the rest of the day, I’ll say “let it go”. Not
knowing the precise location of your data is a positive, both from an
availability and security perspective. You may not know precisely where your
word doc is, but neither does anyone else and, more to the point, in the age of
error coded data and 2048bit encryption, it doesn’t actually matter. Being able
to access that word document anywhere and in the face of some fairly high end
disaster situations is more than a fair compromise for knowing exactly where it
is at any given moment.
[1] A disaster can take many forms and mean many different things to
the individual and the business but in general a flood fire or earthquake would
have a similar impact on both in terms of data loss.
[2] See this related object oriented storage blog
