In a recent post on the AWS Compute Blog, Amazon announced a new pricing model for its EC2 Spot Instance offering, allowing users to easily purchase spare EC2 instances for only a fraction of what the same instances would cost if they were purchased on-demand.
Amazon EC2 Spot Instances allow users to purchase spare computing capacity that AWS has available and are up to 90 percent cheaper than a launching traditional on-demand EC2 instances. Because Spot Instances make use of extra AWS computing capacity, Amazon can temporarily pause them, if needed, in order to reclaim the extra capacity for other customers.
While this may not be ideal for processes that require 100 percent uptime, it can prove useful for those who are looking for a simple boost in performance or for applications where there is some flexibility when it comes to fault-tolerance.
The previous EC2 Spot Instance pricing model involved users bidding for available instances, with instances being terminated based on the number of bids that each instance had. The price for a Spot Instance would then change to whatever the highest unfulfilled bid was. This lead to Spot Instance pricing being quite volatile at times.
However, the new Spot Instance pricing model makes it easier for users to know what they will be paying to create a Spot Instance, with prices changing only once per one-hour period. This means that users no longer have to analyze the prices over time in order to figure out when they might be able to get the best deal.
And for those users who would like to put a cap on how much they are willing to pay for spare AWS compute capacity, there is an option to set a maximum desired price from within the AWS console.
What’s more is that the new pricing model has led to a reduction in the number of interruptions that users face when working with EC2 Spot Instances. Because there is no longer a bidding process, users no longer have to worry about being outbid for a Spot Instance. This would sometimes lead to potential interruptions in service as the spare AWS capacity would be given to a higher bidder instead. Amazon mentioned that on average there was an interruption rate of less than five percent in the 30 days prior to its blog post announcing this new pricing model, which isn’t bad given the significant price discount when compared to on-demand EC2 instances.
While Amazon offers a variety of affordable on-demand cloud compute options that provide a great deal of computing power, EC2 Spot Instances can prove useful for applications that may not require a high degree of uptime, as well as for applications or processes that simply require some additional computing power. And with the low price of such an instance, spinning one up when needed, even if it may be temporarily paused every once in a while, could definitely prove worth for a variety of applications.