If there is one burning question for cloud adoption, it is how the maths works out.
I suppose that is the ultimate backhanded compliment, because nobody seems to doubt the reliability, the speed, the feature-set and all those other things that used to bug the hell out of everybody. All that remains is worrying about the cost – or that’s how it feels when you sit in rather too many cloud presentations over too many years, with guys who dare not backtrack on their early aggressive predictions.
What’s really needed is a clear case study – and the word “clear” here has a bunch of baggage rattling along with it. At Cloud Pro we are inundated with "just so" stories, of tiny parts of megacorporates which have achieved a cloud project in incredibly short times with unbelievably huge profits, zero overheads and of course a perfectly budgeted and delivered project plan – absolutely none of which ties up to the classical world of messy, hurried, imperfect, unpredictable, faith-driven IT decision making as found everywhere else.
I’m sorry if this sounds unduly bitter but I earn my bread and butter looking at other people’s projects and sorting out their problems. There could not be a greater disconnect between the life I lead from day to day and the world of cloud marketing. The charmed existence of easy cloud solutions can look like a very distant dream indeed when dealing with spending limits, aspirations, and "bloke down the pub" level advice.
It was a great relief to see some of the statistics coming from Parallels at their summit event in February of this year, because their ecosystem of customers (all web-hosters) deal with the same types of clients I see - the pressures of being small were very much on its mind.
Take one my clients as an example.About a year or so ago I whizzed up to the outskirts of Liverpool to meet up with the team at Hattons Model Railways, the leading player in its sector, with forward order books that can last for over a year and an entirely in-house, aggressively flexible and very fast-moving SQL development team. In this respect it is very close to Microsoft’s ideal case study, since it seems that there’s far more internal development of this kind in the US than there is in the UK.
Hattons was permanently being hamstrung by internet bandwidth. I suspect if it were a more mainstream enterprise then BT would have figured out it was worth supporting – but despite all its heavyweight software development chops, it was always concerned by the what-if of BT mucking up its business grade link simply because it was in an exchange surrounded by consumer-grade links. This never actually happened while I was in regular contact, but I can entirely understand the concerns: a three-day outage for a consumer is pretty annoying, for a business with millions of pounds composed of thousands of tiny transactions, it would be catastrophic.
So this is a top quality case study, because I am able to report very clearly on how it was "before" – I wrote a business clinic piece for PC Pro on how it could at least partly address its reliability woes. A couple of weeks ago, I got a summary of the “after” situation – they had looked for a while at some hypervisor and virtualisation options (including a little-known player called ProxMoix). Eventually the company settled on a complete suite of VMs and service platforms for databases and web servers, all within Microsoft Azure.
Now, this is not a neat and tidy deployment – it’s a notably ”impure” solution. Hattons can’t quite do everything it wants just by having a plain HTTP server or a plain SQL resource with its existing code base, or by transplanting its existing servers across to Azure Infrastructure VMS. Instead, the company has done a mix and match deployment. Some services are on VMs that look no different to the admin than the physical machines they replaced; others are just in service layers that exist in the Azure Control panel, but don’t have a desktop with a traditional remote-control login at all.
Hatton’s IT manger, Graeme Hogg, is very pleased with the improvement to his peace of mind. The main thrust of Microsoft’s marketing for Azure - that his deployment can scale up and down in response to demand - doesn’t especially seem to bother him. What he likes is the idea that all the basic recovery processes and avoidance-of-mistakes work when it comes to service delivery is out of his hands – and since these guys were previously dependent on their own server hosted in a bare rack hosting centre, with all the security-management and server update duties that implies. Now, by only using full scale server VMs where specifically needed, Hattons gets the world-sized defence capabilities of a web and e-commerce business thousands of times its size.
For how much, I know you are begging to ask? So far, a few months in, the monthly figure is hitting a £1,000: so figure on an annual budget of about £15,000.
There are many things we can say about this project but for me two basic points stand right out. The first is that Hattons in no way came into this project as walking wounded. It was not so much dissatisfied with its old Windows servers, as concerned by the Business Continuity implications of only having one of each. Hattons is also not typical of British businesses, having almost no third-party software products to worry about or external developer/supplier nerves to smooth. If Azure required some basic coding work, it was able to do it itself, and exquisitely familiar with the systems it was dealing with when a change or a re-architecture came up.
The second is that I suspect Hattons might have made quite a different decision only a few months later – and it’s one it may yet make. Server 2012 R2 is such a leap forward if you have been living with 2003 or 2008 that it takes some time for all the new stuff to fully sink in, lots of which a) massively increases system speed on even older servers and b) extends the inter-product links found on Azure to the local server, via tools like the Azure Pack. I would like to bet that Hattons could bring its Azure bill down by 40-60 per cent with even the humblest of local servers equipped with Server 2012 R2. One of the real beauties of the Azure approach is that this option has not been closed off to them.
Do I expect this example to be a sign of plain sailing for everyone else who might read this piece? No, not at all, because I have seen enough of both its setup, and the state of most other networks, to know that a large part of the easy ride into cloud came because it was already leading the field when it came to a clean, fast, well-maintained setup. People with broken networks are quite likely to end up with broken clouds too. Jumping to cloud as a way of tidying up years of bad decisions and bad advice is not the sort of approach that delivers good, clean, capable IT.
