I’ve written about a few of the talks from the recent Unvirtual unconference, including Tim Moreton’s look at big data in the cloud; and Justin McCormack’s ideas for re-architecting cloud applications. I’ve also written previously about a previous Joe Baguley talk on why the consumerisation of IT is nothing to do with iPads.  The last lightning talk that I want to plagiarise (actually, I did ask all of the authors before writing about their talks!) is Simon Gallagher (@vinf_net)’s “notes from the field” talk on his experiences of implementing cloud infrastructure.

Understanding cloud isn’t about Amazon or Google

There is a lot happening in the private cloud space and hybrid clouds are a useful model too because not everybody is a web 2.0 start-up with a green-field approach and enterprises still want some of the capabilities offered by cloud technologies.

Simon suggests that private cloud is really just traditional infrastructure, with added automation/chargeback… for now. He sees technology from the public cloud filtering down to the private space (and hybrid is the reality for the short/medium term for any sizeable organisation with a legacy application set).

The real cloud wins for the enterprise are self-service, automation and chargeback, not burst/flex models.

There are three types of people that want cloud… and the one who doesn’t

Simon’s three types are:

1. The boss from the Dilbert cartoons who has read a few too many analyst reports (…and is it done yet?)
2. Smart techies who see cloud as a springboard to the nest stage of their career (something new and interesting)
3. Those who want to make an inherited mess somebody else’s problem

There is another type of person who doesn’t welcome cloud computing – people whose jobs become commoditised.  I’ve been there – most of us have – but commodisiation is a fact of life and it’s important to act like the smart guys in the paragraph above, embracing change, learning and developing new skills, rather than viewing the end of the world as nigh.

Then there are the politics

The first way to cast doubt on a cloud project is to tell everyone it’s insecure, right?

But:

• We trust our WAN provider’s MPLS cloud.
• We trust our mail malware gateways (e.g. MessageLabs).
• We trust our managed service provders staff.
• We trust the third party tape collection services.
• We trust out VPNs over the Internet.
• We may already share datacentres with competitors.

We trust these services because we have technical and audit controls… the same goes for cloud services.

So, I just buy cloud products and start using them, yeah?

Cloud infrastructure is not about boxed products.  There is no “one-size fits all” Next, Next, Next, Finish wizard but there are complex issues of people, process, technology, integration and operations.

It’s about applications, not infrastructure

Applications will evolve to leverage PaaS models and next-generation cloud architectures. Legacy applications will remain legacy – they can be contained by the cloud but not improved. Simple provisioning needs automation, coding, APIs. Meanwhile, we can allow self-service but it’s important to maintain control (we need to make sure that services are de-provisioned too).

Amazon is really inexpensive – and you want how much?

If you think you can build a private cloud (or get someone else to build you a bespoke cloud) for the prices charged by Amazon et al, you’re in for a shock. Off the shelf means economised of scale and, conversely, bespoke does not come cheap. Ultimately, large cloud providers diversify their risks (not everyone is using the service fully at any one time) and somebody is paying.

Opexification

There’s a lot of talk about the move from capital expenditure to operating expenditure (OpEx-ification) but accounts don’t like variable costs. And cloud pricing is a rate card – it’s certainly not open book!

Meanwhile, the sales model is based on purchase of commercial software (enterprises still don’t use open source extensively) and, whilst the public cloud implies ability to flex up/down, private cloud can’t do this (you can’t take your servers back and say “we don’t need them this month”). It’s the same for software with sales teams concentrating on license sales, rather than subscriptions.

In summary

Simon wrapped up by highlighting that, whilst the public cloud has its advantages, private and hybrid clouds are big opportunities today.

Successful implementation relies on:

• Motivated people
• A pragmatic approach to politics
• Understand what you want (and how much you can pay for it)

Above all, Simon’s conclusion was that your mileage my vary!