Spot Instances - Increased Control
Today we announced the launch of an exciting new feature that will significantly increase your control over your Amazon EC2 Spot instances. With this change, we will improve the granularity of pricing information you receive by introducing a Spot Instance price per Availability Zone rather than a Spot Instance price per Region.
Spot Instances enable you to bid on unused Amazon EC2 capacity. Customers whose bids exceed the Spot price gain access to the available Spot Instances and run as long as the bid exceeds the Spot Price. Spot Instances are ideal for use cases like web and data crawling, financial analysis, grid computing, media transcoding, scientific research, and batch processing.
Since the launch of Spot Instances in December 2009, we have heard from a large number of customers, like Scribd, University of Melbourne/University of Barcelona, Numerate, DNAnexus, and Litmus that Spot is a great way of reducing their bills with reported savings as high as 66%. As a part of that process, we also realized that there were a number of latency sensitive or location specific use cases like Hadoop, HPC, and testing that would be ideal for Spot. However, customers with these use cases need a way to more easily and reliably target Availability Zones.
As defined, Spot Instances allow customers to bid on unused Amazon EC2 capacity and run those instances for as long as their bid exceeds the current Spot Price. The challenge with these use cases was that Spot Instances were priced based on supply and demand for a Region, however Spot Instance capacity was constrained per Availability Zone. So, if we wanted to fulfill a request targeted at a particular Availability Zone, we could potentially have to interrupt more than one instance in multiple other Availability Zones.
Rather than causing volatility in the market, we are announcing a shift in the Spot pricing methodology. Now Spot Instance prices will be based on the supply and demand for a specific Availability Zone. By shifting the unit of capacity we are pricing against, customers bidding strategy will directly determine whether or not they are fulfilled. Existing customers who do not target their bids will have their instances automatically launched into the lowest priced Availability Zone available at the time the bid is fulfilled.
We have already seen customers successfully run HPC workloads, Hadoop-based jobs (as shown in the BackType case study), and testing simulations (as shown in the BrowserMob case study) on Spot. However, we are hopeful that this change will make it even easier for newer customers to immediately start capturing significant savings. Moreover, this change blazes the way for us to introduce Spot integration with Elastic Map Reduce.