Friday, 15 February 2013

99.999% Availability - Really

So… How many times have you heard it / seen it… Vendor A walks into the business that your IT Team supports and says… Yes –  of course our product is highly available… We guarantee it… 99.999% available…. all good… the “five nines” have been quoted –  what could possible go wrong..

The business then goes and signs on the dotted line (ok –  see anything wrong here –  yep –  the business went and done it again, they let a vendor sell into the business without involving their infrastructure teams… but hey –  its still all good –  Vendor A has stated that Product B is 99.999% available –  its the best possible choice for that mission critical business application) –  What could possibly go wrong

So many months later after the business has brought this wonderful piece of infrastructure and started running their business critical app on said kit –  it suddenly looks a little more difficult to achieve this system integration and the “five nines” promise… why well –  erm that patch management that needed to be completed wasn't really included in the quote of system availability and yeah, the lifecycle management activities that need to be completed also means “planned downtime” –  surely this eats into out 99.999% availability as well –  Yep pretty much… so what should we learn from this –  probably a couple of things when dealing with vendors and guarantees of availability;

  • Ensure that the business involves infrastructure at a very early stage –  so vendor cannot mislead with quotes of availability
  • Identify your availability requirements ahead of engaging vendors –  at least understand your availbility requirements and work back from there with the vendor
  • Be sceptical and dont believe the hype!
  • Understand what is quoted in this 99.999% availability i.e. does this include planned and un-planned ourages
  • Understand patch management regimes and what this may mean to your availbility characteristics
  • Understand lifecycle management application and possible implications on downtime whilst running these processes
  • What currency practices does the vendor have in place to ensure your equipment stays in support and what is the frequency of updates that will be required
  • Understand the vendors measurement of availability –  i.e. just because a storage array is up and running and servicing I/O's slowly does this count as data available or unavailable
  • Understand the penalty system that a vendor puts in place for not achieving their guaranteed availability and how this is invoked –  it always helps to motivate a vendor with financial penalties when availability is not being met
  • Put proper measurements in place for your infrastructure –  when you hold your monthly / yearly vendor updates you can challenge them with appropriate metrics. This again drives good behaviour for us (the consumer)

@storagebod has written a great article on this topic also –  and can be found here:

Hopefully some food for thought!





Sunday, 10 February 2013

Stack Wars - Is 2013 the year of the stack???

There has been quite a lot going on out there by various vendors, a whole bunch of them trying to demonstrate how you can ship entire pre-loaded infrastructure stacks (i.e. storage, network, compute, memory, hypervisor (In some cases) and OS) into shippable rack configuration, calling them virtual data center's (VDC) and taking away the need to do some of the typical stuff that IT departments have to do for their customer base… Examples of these products would be:

  • VCE vBlock
  • HP Matrix
  • FlexPod (although would not put this in the exact same category)
  • IBM PureSystems

Additional to this, other vendors have also been making inroads into going further than just the core infrastructure layer and have started offering application services, examples of this could be:

  • Oracle Exadata (offering database services) / Oracle Exalogic (going into the middleware space, OBIE, OBIA, SOA etc…)
  • Teradata database machine
  • IBM PureApp (part the pure range) –  incl websphere, UDB DB2, BPM etc
  • IBM Z-Series / Mainframe (it really does provide the whole stack –  but is pretty bloody expensive!!!)

Some of the offerings have been better than others and i also think its fair to say as a general statement –  that a level of maturity just hasn't been there in the past…But… Are we starting to see some light in this space???

So firstly a couple of observations on the infrastructure stacks (the top list i mention) – Having observed a number of engagements on some of these products (either personally or through acquaintances) over the last 24 months –  in the early days these installations, ideas of just bringing in the infrastructure –  putting power and network up the backside of said device and consuming it were just not there –  however, lessons have been learnt and definitely for green-field sites or areas where legacy infrastructure doesn't hinder progression –  some vast improvement to product and ability to deliver has started to be obvious…. There is still a large amount of integration activity to take place, OS builds and standards to be engineered but the promise of easier delivery etc is starting to emerge with some promise. I have to say, whilst i was no real big fan of VCE –  vBlock 12–18 months ago –  it really seems to have taken shape

Now –  onto the end-to-end stacks –  the true stacks, where the vendor isn't just providing the infrastructure but also including application ready for business logic / app to be applied.

Now –  when you take a look at the vendors involved Oracle should really have this one killed… They have the exa range, they have the complete set of business applications to consume and if you believe the website, marketing –  its all ready to rock and roll with little to no integration required to start using the thing –  wheel it in, switch it on and start consuming…. oops –  doesn't quite work like that. It has to be said –  that Oracle really do have all of the components needed to make a holistic business offering –  but it just isnt quite there yet… it will get there, it needs a few more cycles of maturity before we can take full advantage…

Teradata –  guess they have a good(ish) database appliance that is meant to be efficient for large sets of data where multiple business lines may consume same data set –  but sort of getting a bit long in the tooth. DR Story is questionable and longevity of the company… hmmmm lets wait and see

IBM Mainframe –  OK –  i couldn't get past mentioning this somewhere in this article –  but its got to be said, in terms of someone who owns the stack, can layer OS, business app, database etc etc –  the Mainframe has had years on all of the other open systems stuff that is still trying to grow up –  admittedly its sort of expensive –  but i guess my point is, dont overlook it –  there is a lot of business value to be added / derived from potentially utilising spare cycles / mips on your big iron –  its prob worth taking a look… and whilst on IBM –  I guess i move onto my final nomination –  IBM PureSystems / PureApp

IBM Pure –  so have been watching this / following this for some time and also listening to some of my industry peers experiences of this product set –  and its got to be said –  its looking good. If you happen to be a IBM house, and someone that utilises the IBM Product set (websphere, BPM, DB2 etc etc) then you could do worse than take a look at pure-systems product range. You get a full environment in a box, all nicely integrated (so it appears) with some pretty smart tools wrapped around it. You can run P or X86 blades so any existing custom code you may have written can continue to be used. As a stack that you can wheel in and start using –  it looks pretty tight. In my humble opinion IBM took all of the integration lessons that they learnt with the mainframe environments and went an applied it to a distributed systems stack –  and with a pretty damn good result.. So IBM looks like they have come from middle of the pack to a clear first…

So will 2013 become the year of the stack – Probably not everywhere but the market is definitely starting to mature and IBM may just have put themselves back on top!!