This startup can run apps on Amazon’s cloud for 1/10th of the regular price

The engineers at startup ClusterK know a heck of a lot about an often overlooked part of popular cloud provider Amazon Web Services called Spot Instances, where you can name your price for using slices of servers in the Amazon cloud. ClusterK has figured out a way to run complex applications on these Spot Instances for a tiny fraction of the cost of running it on Amazon’s more popular on-demand instances.

Dmitry Pushkarev, ClusterK’s chief executive, told VentureBeat in an interview that his startup can give companies “90 percent off the Amazon price.”

[aditude-amp id="flyingcarpet" targeting='{"env":"staging","page_type":"article","post_id":1664676,"post_type":"exclusive","post_chan":"none","tags":null,"ai":false,"category":"none","all_categories":"big-data,business,","session":"C"}']

Which is huge. Companies are already keen on lowering their bills when it comes to running their applications in the public cloud, whether they’re for internal use or external consumption. The cloud price wars, where Amazon, Google, and Microsoft, and other companies are challenging each other, have made this clear, and optimizations with spot instances could be considered a new way to achieve that nirvana.

Companies that have heard about this tool absolutely love it. And yes, investors are funding it.

The startup has now raised $1.2 million from Rally Ventures, Data Collective, Mayo Foundation for Medical Education and Research (as in the Mayo Clinic), and several angel investors. (Disclosure: Rally Ventures is one of the investors in VentureBeat as well.)

ClusterK’s system predicts prices for different types of Amazon Spot Instances and then it distributes capacity for computing work in such a way that computations are simultaneously cheap and fault-tolerant.

“So when those [price] spikes do occur, in most cases, they won’t affect our capacity,” Pushkarev said. “We’ll try to rebalance that well ahead of time.”

What’s more, companies don’t need to tinker with their code if they move applications from on-demand instances on Amazon to its Spot Instances, Pushkarev said.

ClusterK’s team members talk with people on the Amazon’s Spot Instances team every day, Pushkarev said. And not only that — he said the people at Amazon love ClusterK. The startup is increasing adoption of Amazon’s Spot Instances.

The question is whether Amazon might end up competing with ClusterK at some point in the future — just as it has come out with products that compete with products from customers like Dropbox and Parse. But Pushkarev doesn’t think that will happen for one to three years, he said. So all is well.

[aditude-amp id="medium1" targeting='{"env":"staging","page_type":"article","post_id":1664676,"post_type":"exclusive","post_chan":"none","tags":null,"ai":false,"category":"none","all_categories":"big-data,business,","session":"C"}']

In any case, should other public clouds come out with their own versions of spot instances, ClusterK could well support them. Amazon would be unlikely to do that.

Before starting ClusterK, Pushkarev in 2011 cofounded a startup called Moleculo, which developed new ways to sequence DNA using Spot Instances on the Amazon cloud. “It got the price down from $400 per sample to $40,” Pushkarev said. Sure enough, biotechnology company Illumina bought it in 2012. Looking back, Pushkarev suspects that the startup saved around $1 million in computing time.

The success caused Pushkarev to wonder what he might be able to do for other types of businesses.

And so in 2013 Pushkarev left Illumina and started ClusterK. Now headquartered in Palo Alto, Calif., ClusterK has attracted a handful of customers. Pushkarev pulled up a dashboard showing one customer’s Amazon Spot Instance costs and savings. The savings relative to the costs were impressive. One day earlier this week, the customer saved over $1,000 running Hive queries on its custom Hadoop cluster, Pushkarev said.

[aditude-amp id="medium2" targeting='{"env":"staging","page_type":"article","post_id":1664676,"post_type":"exclusive","post_chan":"none","tags":null,"ai":false,"category":"none","all_categories":"big-data,business,","session":"C"}']

“Every day on just a small Hadoop stack, they’re saving over a thousand dollars, which allows them to run more experiments, get results faster and stay ahead of the competition,” he said.