ClusterK Getting The Most Out Of Amazon Cloud, Reduces Apps-Running Price Up To 1/10th
Cloud computing has gained its fair share of using among some major tech companies world wide. Amazon Web Services product, the Amazon EC2 (Elastic Compute Cloud), became a favorite in providing efficient, secure and scalable cloud computing resources for many business needs like Analytics, Financial Modeling, Big Data, High quality graphics rendering, etc.
Cloud computing services usually do come up with hefty price tag and likewise Amazon EC2 solutions are no exception. Now comes the enchanting role of ClusterK, a startup who have best utilized the low-cost alternative (Amazon EC2 Spot Instances) to Amazon EC2 services, as the prices for Spot Instances are significantly less than the On-Demand pricing of Amazon EC2 instances.
The motivation behind Amazon EC2 Spot Instances was to offer cheaper solutions for non time-critical and fault-tolerant tasks. Spot Instances lets users bid for spare Amazon EC2 instances. And if their (user’s) bid exceeds that of the spot price, set by Amazon EC2, the instance resources (servers) are assigned to the applicant till the time they terminate their usage or the Spot Price exceeds their bid price.
Engineers at ClusterK have come up with algorithms that let their clients run the most complex of applications on the servers acquired through Amazon Spot Instances. The system functions in a way that it predicts bidding prices for the Spot Instances for their clients. After the bid succeeds the implementation seamlessly distributes compute capacity and resources in an error-free way.
Furthermore, ClusterK also has devised special routines which let them retain Spot Instances. The routine is preemptive in nature as it predicts likely highs and lows in the real time bid rates. As CEO Pushkarev puts it:
“So when those [price] spikes do occur, in most cases, they won’t affect our capacity, We’ll try to rebalance that well ahead of time.”
Another important functionality provided to by ClusteK is that the clients would not to bother refactoring their application’s code if they’re transferring work load from on-demand instances on Amazon to its Spot Instances.
Not only current and prospective clients applaud ClusterK’s efforts, Amazon too love this initiative, said Pushkarev. Adding to that, he said his team is constantly in touch with the people at the Amazon end, implying a healthy collaboration.
As written before, ClusterK’s efforts are not just welcomed but rewarded as well. They’ve managed to raise approximately $1.2 million from investors like Rally Ventures, Data Collective, Mayo Foundation for Medical Education and Research (as in the Mayo Clinic), and several others.
After sometime Amazon might end up considering ClusterK a competition but as Pushkarev said, this might not happen within the next 2-3 years because money they are saving is not just for the sake of profiteering but also to invest in research so that they could support any other public clouds coming up with their own version of Spot Instances.