What your company knows is its most valuable asset, so don't let
that knowledge data be put in some dark, isolated place.
Storage is often regarded as somewhere to keep data in between
processing it. In the days when your choice was IBM or a
plug-compatible manufacturer, the storage contract was a
consolation prize to the company that lost out on the processor
bid. Those suppliers that sold both focused on their processors and
treated their storage products as the Cinderella of their range.
But just as Cinderella had the last laugh, the significance of
storage solutions is changing.
System heterogeneity, despite what the pundits say, is not a virtue
- and having a large heterogeneous network is actually a sign of
management failure. At times organisations go out of their way to
install as many diverse platforms as possible, thus erecting
artificial walls between their applications so that they can
occasionally crow about leaping over one or two of them and
achieving some fleeting synergy.
When mainframes were kept in glass rooms, application synergy was
assumed, since a common character set, file format, programming
language and environment for all applications rendered all data
available for multiple use.
That has been lost largely, for no good technical reason but
because of supplier politics and short-term strategies. Although
today's mainframes fit in shoeboxes and their operating system,
typically MVS, will run on a sub-1kg laptop, artificial barriers -
such as software pricing - remain.
But the data on all these disparate systems is valuable. It is said
that a company consists of its employees, but it really consists of
their knowledge. Recruitment adverts still stress experience over
skill and are veiled attempts to buy more knowledge. If what a
company's employees know is valuable, how much more important is
what the firm knows? As long, that is, as this knowledge is
available. For data to be available to many platforms, it has to be
in a form they can all understand, which may be a form native to
none of them or, in a word, virtualised.
The good thing about standards is that there are so many of them to
choose from. The same is true of virtualisation concepts and they
are just as diverse. What the term means to any given supplier
depends largely on that supplier's starting point in the market,
but to the user the differences can be crucial. The pooling of
available space and controlled assignment is sometimes called
virtualisation, but it is really not even management -
administration is a better term. Similarly solutions tied to
specific subsystems or processors have very limited futures. A
network-based approach is the only one offering open-ended
functionality.
For the ideal storage solution, the primary requirement will always
be reliability in all its facets: availability, security, disaster
recovery, and so on. Efficient resource management becomes less
important as technology gets cheaper and autonomously managed. But
the ability to share data increases in importance. Sharing can be
at the subsystem level with a proportion accessed from one platform
and a proportion from another; at finer granularity but still
screened from other systems; and with one platform owning a space
and another being allowed to read it.
Data sharing - at a record or even data element level - is a long
way in the future but has to be the eventual goal. In this context,
true data sharing is an incredibly complex issue. If a logical
database is distributed across many platform types, who does the
locking and change logging? Even worse, how are transaction
back-out and point-in-time restores handled across multiple
platforms with many applications to be notified?
In-band metadata-based virtualisation, coupled with self-describing
data, has the potential to restore homogeneity and therefore
application synergy, while supporting freedom of platform choice
and network-level availability. Not every supplier's current
roadmap holds the promise of ever getting there.
Phil Payne is principal at Isham Research