Thursday, 9 September 2010

Cloud services - As long as its x86 only!

The current release of cloud enabling products, automation services,
virtualisation products and hardware stacks that are allowing us to
build internal and external cloud services at the moment is stunning.

There is a whole heap of acquisitions that are occurring, VMWare bring
products such as vCloud Director to bear and spring source for the
development community are driving things into a bit of frenzy (and i
almost forgot to mention that Microsoft are out there winding it up also).

However - I have a problem with all of this brand new shiny stuff / way
of working - and it is just that - its brand new and shiny, and only
potentially accounts for the net-new environments that are being
provisioned (and not even all of these are being catered for (SPARC?
AIX? Sol Zones?)).

The whole point of cloud services is that it is just that - a service,
and should be able to abstract any platform regardless of what hardware
/ OS / virtualisation product / application that maybe running.

Few of the vendors / end users out there are taking notice of the "to
hard to do pile" and are focusing on the easy to do pile (OS's that run
on x86 and can be virtualised by either VMWare or Hyper-V seem to be the
current trend!)

What about AIX?
What about Solaris on SPARC?
What about being able to span Solaris Zones
and the list continues (if you want my full list - please feel free to
contact me)

As for the argument that commodity hardware is where we going and we
should focus on this area (again - read x86 platforms) we have a huge
current estate that covers a whole bunch of non x-86 platforms and i
would want to do the lot... Surely the vendor community does not want to
suggest i now have two support teams, one for new shiny cloud and one
for current way of working? - That's just not viable!

One last thought on this subject - All vendors are now coming out with
there own vertical stacks (such as oracle exadata) - we will see more of
this trend, be it server, database, application - so these cloud
enabling tools had better be able to intgrate with these type of
approaches or that to-hard pile is going to start increase, and the
whole concept of cloud will just fall away.



1 comment:

  1. To be fair you're already running more than one support team anyway; Wintel in most places has the biggest density and it's going to be the Wintel space that will take the biggest hit if Cloud turns out to be a success.
    The real question is how much is it actually going to cost to recognise benefits which, to a large extent, we could already utilise by better management of our current vm farms and by getting the app teams to write more scalable solutions rather than the instance focus they have currently in a lot of areas.
    At the moment it smacks significantly of shiny new toy syndrome.