Listed on

Head in the Clouds

If you open any computing trade journal you could be forgiven for believing we had all moved to New Zealand and were looking out on a marvellous vista - the land of the long white cloud. Unfortunately, in spite of the great work being done by the marketing people, the world hasn’t actually changed that much. Cloud is still virtualised servers and remote hosting. The new and exciting bits are rapid on demand deployment and tear down. In this article I take a look at who and why you should use cloud from an economic management point of view.
Basic cloud economics is essentially renting a computing resource. It is almost exactly analogous with renting commercial equipment. This insight should allow managers in the non-computing area to easily get a grasp on the costs and benefits and cut through the hype.

Supply side



Is cloud cheaper? To answer this question we can think about under what circumstances could a car leased from an agency be cheaper than buying it:
  • If the agency has access to far cheaper prices than you can get due to a deal with the manufacturer or cutting out middle men
  • Their overheads are sufficiently low that they can make adequate profit without fully absorbing any buyer advantage
Do these conditions hold true for cloud providers?
  • The computer hardware market is highly competitive, commoditised and has razor thin margins - so there is almost no buyer advantage to be had on the hardware
  • The electricity market, on the other hand, has quite a wide range of price points and highly efficient provision of cooling by large well run datacentres can provide significant price advantage for a cloud provider. The gap is less wide than the naive price difference as equipment owners can rent space in datacentres with good buying power and at least good cooling ratios (though it has to be admitted that the large cloud providers still remain ahead)
In the wash up it appears that cloud CPU cycles are a bit more expensive than owning your own gear on a large scale.

Demand side



From an economic point of view who should use cloud? Just like the equipment rental market there are many circumstances under which it pays to rent rather than buy:
  • Limited capital - if you can’t afford to buy but have a good business proposition then rental will let you get started without having to front the purchase capital.
  • Limited need - if you don’t need the equipment full time why buy it. The economic decision point is trivial to determine - if after adding in your overheads the cost of renting is less than buying, then rent it.
  • Limited skill - like rental equipment needs servicing, maintaining the infrastructure cloud servers need requires skilled technicians - it may not pay to acquire the skills required to manage the environment in which the equipment is kept and hence be worth a premium to look after the servicing

Also like equipment rental, you might choose to own some equipment and rent more at other times. For instance, when you have extra work or during expansion. In this case private cloud makes particular sense as this enables you to easily take advantage of cloud facilities when you need them - this is the same concept as buying equipment that is compatible with that used by your rental provider so you can use theirs without a lot of additional training.

Conclusion



Cutting through the hype of the cloud - it makes sense to buy in the cloud:
  • if you can’t afford your own equipment in a data centre with good buying power for electricity
  • if your business needs to cope with increased work for some of the time
  • if your business is growing and you need to meet a need now and later backfill the hardware
  • if you lack the skills to run the hardware yourself and it isn’t worth buying them in
Owning your own equipment is sensible if you have access to the skills to maintain it, have a reasonably constant, predictable workload and access to reasonable electricity pricing with efficient cooling. It makes sense to consider private cloud:
  • if you have changing workloads and need to be ready to expand your processing
Basic cloud really isn’t hard to understand and isn’t a new management problem1. Of course, once you have an easily deployed resource that you can devolve management from a central service bureau to departments2, you can make your own management problems, but that is beyond the scope of this article.




1. The exciting part from an economic perspective is the potential of off-peak processing i.e. servers which become available at less than the cost of ownership because the cloud provider has unsold capacity - but this is once again a familiar management problem exactly analogous to spot rates in the electricity market.

2. Many cloud infrastructures support the ability to devolve management from a central control to others and allow them to bring up and shutdown their own infrastructure elements. This facility could be both a great advantage or a terrible management problem depending on how skills are distributed through an organisation and how the distributed facility is used.