The Airbnb of Cloud Platforms

Spotinst offers a solution that ensures business continuity and reducing costs by AWS. The software runs multiple prediction algorithms along with real-time decision making to determine the most cost-effective slot for the customer’s application

The Airbnb of Cloud Platforms

Cloud providers possess large quantities of spare capacity, and they must either incentivize clients to purchase that capacity or suffer losses.

Amazon Web Services was one of first cloud providers to address this challenge, by allowing clients to bid on its spare capacity. This price model is called “Spot.” Spot is characterized by granting resources to bidders when their bids exceed a periodically changing spot price. By using the Spot pricing model, customers can save up to 90% off the original price. Some say that Spot is the cheapest computing power on earth.

Google Cloud Platform also sells unused capacity in its “Preemptible VMs” market, where buyers can get computing resources for a limited amount of time, up to 70% off the original price.

Other public cloud providers (such as IBM SoftLayer and Microsoft Azure) would naturally like to sell their unused cloud capacity too. They and others are working on similar models to incentivize clients to purchase their unused cloud capacity.

On the other side of the equation, some cloud customers are reluctant to use Spot this pricing model because of perceived dangers inherent in this pricing model, such as when customers must design their applications to counteract situations where the cloud resources might disappear at any moment.

These challenges, along with some existing limitations in AWS Auto Scaling Groups, led Spotinst's founders to conduct research to crack the formula and help customers run complex workloads on top of Spot instances and to save significant amounts of money.

After a series of successful product tests with leading companies in the U.S. and Israel, Spotinst has publicly launched its "Elastitude" product.

The simple, yet key, insight underlying Spotinst is that it runs multiple prediction algorithms along with real-time decision making to determine the most cost-effective slot for the customer’s application, and the best server type at any given moment.

How it works? A set of mathematical tools, an auto recovering system, and a self-learning component do the job. It's like solving a kind of "maxima and minima" function.

An optimizer software runs behind the scenes and digests AWS' historical pricing data. This data shows the amount of available resources that exist in AWS at any given time.

Cross-checking historical information with real time pricing data allows for accurate and intelligent selection of a server.

For example, the software knows to consider pricing anomalies, whereas a specific instance type from a data center may be running "too many" hours with a price that is higher than it should be. Spotinst’s software eliminates this instance and will not select it. 

For example, an instance with abnormal spikes in prices will be ranked with a low score, even though it has a cheaper price than normal.

Whenever an instance suddenly becomes not profitable, the software will recognize it and will gradually replace the instances to ensure maximum cost savings. This will spin "better" capacity from other sources.

It is getting even more complicated when parts of the instances have 2 cores, and others have 4 or 8 cores.  Spotinst rebalance the traffic offensively to match the application needs across the cluster.

You might be interested also