To me the private cloud is an anti-cloud – I don’t see how a private cloud can possibly provide the economies of scale and office empire neutraility to make it an attractive alternative to a standard heavy-metal data centre. But maybe, just maybe, the adoption of the private cloud will begin to pave the way, in terms of attitudes and technologies, for the real cloud.
Consider a simple business resource – office parking. Since parking spots are all a standard size and that they are either empty or not, you would think that they would be a good candidate for automatic allocation that would be able to accommodate the needs of the business (based on a calculation of how many parking spots are required per square metre of office). For some reason, not matter the intention, office parking is a maze of complex interdepartmental relationships, interpersonal relationships, politics, favours, back-stabbing, grease-pole climbing and cronyism. Something that a technical person would understand as being simple becomes very complex very fast. The end result is that parking is meticulously allocated yet under-utilized and over-subscribed where visitors and contract staff are left wondering why they have to park in ‘overflow parking’ a few blocks away in a parking garage with ‘Danger Flooding’ signs while the pristine office parking, where they can’t get a space, lies partially empty like the trendy City lunch hangouts are these days.
It has been a few months since I had much to do with a large data centre, but in my experience the resources that should be reasonably easily to allocate are, in reality, not easy to allocate at all. SANs, which were supposed to give easy access to storage in a ‘private cloud storage’ way, do not live up to the ease of allocation. It seems that it takes months to get storage, the data centre is always running out, the configuration of the network is all wrong or some other excuse that drags out the process and heaps up the bills. But if you know the right people to speak to you can get storage allocated in the same day as it is ‘borrowed’ from somewhere else. I have seen memory for blade servers also being ‘borrowed’ in this way as well as ports on switches or even IP addresses becoming a commodity that is traded between projects – ‘I’ll swap 500GB on the SAN for a 1GB port on the switch, deal?’. Standard servers ordered for particular projects take ages to provisioned as they become re-allocated to more high-profile projects as soon as they arrive. Costs are piled on top of the original price as vendors throw complex licensing schemes at the configuration where the backup software for this particular database on this particular SAN using some other particular switch is the only one available and costs a fortune.
Maybe I am being cynical but I find that large enterprises are mostly incapable of keeping IT infrastructure costs under control and it is not uncommon that the infrastructure costs more than the software development effort. That is what shocks me the most - where a team of very smart technical people pour their souls into a project and find that their costs, and value, is minimal compared to the infrastructure costs. One would think that by now we would have progressed to the efficient usage of internal IT infrastructure resources but, even though over the years we have moved from physical to virtual hardware, the promised efficiencies have not materialised. The blame does not necessarily reside solely with IT and the rate at which business is demanding new solutions, storing and analysing large amounts of data, and being unable to accurately predict their infrastructure needs places additional strain on those trying to run a data centre properly.
The attractiveness of the cloud is that the promised goal of being able to provide just enough computing resources on demand by entering into a contract and SLA directly with the cloud provider can be achieved – effectively bypassing the internal IT ‘game of silly buggers’. In addition the cloud, by having the economies of scale, will be more cost effective than internal IT. So you have a combination of the best processes to provision equipment and reduced cost – so surely the decision is a no-brainer? Unfortunately the cloud is still in its infancy and there are some problems that need to be sorted out. There is the issue of data security, regulatory requirements and so on, there are still some question marks around SLA’s, accountability and who can be sued, but the biggest, bar far, is the change management needed in internal IT. In house infrastructure and data centre teams need to move beyond setting up of individual servers, storage devices and other equipment and move into a word where infrastructure can be assigned and allocated at the click of a mouse from a dashboard.
So the concept of a private cloud, while still a reasonably expensive and finite IT resource, still has a part to play in the changing of the way data centres and businesses utilise their computing resources. The technologies for managing resources will develop, the hardware manufacturers, operating systems, backup and restore technologies and networking infrastructure will need to become more cloud capable and more usable in a cloud context. Cloud oriented system architectures, frameworks and runtimes (Platform as Service) may also succumb to the demand to be private – at least for now.
Once that happens, once all the attitudes have changed, once all the technologies have matured and once business has come to terms with the legal and regulatory issues then we will have a situation where an operator can sit at a terminal and drag and drop an application from the local data centre to one at half the cost in Peru. So in achieving that ultimate goal of the real cloud I welcome the private cloud as part of the adoption curve and a driver for the solutions.
Simon Munro @simonmunro