Amazon’s Spot Price Secret
Being able to pay only for what you use on AWS’s EC2 platform has closed the gap between the IT capabilities of large and small companies. Ten years ago the only way to access high-end computing power was to either buy it yourself or lease it – on a long and expensive contract. Nowadays, anyone with a credit card can spin up a server cluster for less than a night out costs.
Here’s an example: a server with 60.5 GB of memory, 88 EC2 Compute Units (2 x Intel Xeon E5-2670, eight-core CPU’s) and 3370 GB of storage will set you back $2.40 an hour. It is truly amazing and has changed the nature of IT delivery forever. Of course Amazon is not the only player in this market but they are the one we use.
Their spot-pricing option makes it even more interesting. Excess capacity is sold to the highest bidder at the current spot price, which is generally lower than the On-Demand price. Your instance will fire up when your bid exceeds the spot price and shutdown when below. Here’s the trick: putting in a really high bid (say three times the on-demand price) will spin up your instances and keep them running pretty much indefinitely – and you still only pay the current spot price. At the time of writing, that same server cluster is firing up at $0.54 per hour on the spot market! Massive savings but you have to be able to plan your architecture for the risk of instances shutting down at any time. A good way is to have a minimum capacity running under the guaranteed on demand or reserved model with spot capacity to supplement that.
When you are offering a level of IT services for free, as most cloud providers do, it’s important to be able to tune your hosting costs to your business model. The same counts for prospecting or POC projects – make sure you can afford to provide service at limited or zero cost, and always ensure that you can scale down as quickly as you scale up.