In this special guest feature, Mike Maciag, COO of Altiscale outlines a formula for working to improve ROI for Hadoop deployments. Mike has a long history of building companies in enterprise software markets. Prior to Altiscale, he served as the president and CEO for DevOps leader Electric Cloud, where he grew the revenue from zero to tens of millions while building a worldwide presence and signing hundreds of blue-chip customers. Mike holds an MBA from Northwestern University’s Kellogg School of Management, and a BS from Santa Clara.
With the adoption of any new technology, businesses are tasked with measuring success and demonstrating value to the organization — Big Data is not any different.
As interest in Big Data grows, the value of the technology platform that delivers it must be accurately calculated and demonstrated to the rest of the organization. Hadoop is increasingly the platform of choice for Big Data, and there is a common misconception that predicting and deriving ROI from Hadoop deployments is too challenging. According to a recent Gartner research report, 24 percent of organizations are not measuring the ROI of Big Data at all.
This lack of measurement can slow the adoption of Big Data and its benefits, since there is no agreed-upon value of the data to the organization. While there is certainly an element of difficulty in predicting and receiving real economic return when transitioning to a new technology like Hadoop, there are also proven strategies to extract the most value.
Whether you’re already leveraging Big Data or are about to deploy it, there are ways to better predict and extract ROI. No matter which situation you are in, you can shift your strategy to find the most value.
Hitting the ROI Wall with DIY Big Data
When an organization is already leveraging Big Data, but has been managing an implementation on its own, either on-premises or with a bare-bones infrastructure cloud provider, they inevitably reach a point where the system stops scaling with ease. Big Data platforms are large, distributed systems with complex components, and scaling them is tough.
Inevitably, teams contend with lengthened job completion times, declining job completion rates, and rising costs in the form of people, hardware or spot instances. Ultimately, data scientists get increasingly involved in operations, spending more time maintaining the Hadoop implementation than doing what they want to do — drawing insights from the data.
When you’ve hit this wall, your financial benefit projections also stall. This is not a sign that your Big Data initiatives won’t pay off. Rather it’s a sign that you need to shift your thinking and, perhaps, the way you run Hadoop. To improve your ROI, look for technologies that help do the following:
- Boost data scientist productivity: Too often organizations are using their expensive, analytical human resources to serve as operations resources to scale Hadoop. Instead, they could employ technology solutions that help mitigate job failures, accelerate job completion times and improve job reliability. By moving to a solution that includes operational support, such as a Hadoop-as-as-Service (HaaS) solution, expensive data science resources are freed up to focus on job results.
- Reduce labor costs: Switching to an infrastructure that is better optimized for Hadoop means operations can run more smoothly and job failures are reduced. This significantly reduces labor costs, specifically the number of Hadoop engineers and operators needed to manage and maintain Hadoop implementations.
- Avoid the costs of wasted compute cycles: By reducing failed jobs, businesses also reduce wasted node provisions and ongoing computational costs.
When deciding to transition to a new form of Hadoop implementation, estimate the quantifiable ROI by recognizing increased data scientist productivity, reduced labor costs and the elimination of costs due to wasted compute cycles.
Predicting Future ROI for Those New to Big Data
If you haven’t started using Hadoop, determining future Hadoop ROI becomes a bit like drilling for oil — it’s hard to know what the results will be until the drilling process has already begun. In other words, it’s difficult to quantify the benefits of Big Data insights when those insights are not yet known. On-premises solutions have high upfront costs and require ongoing support, leaving organizations looking for better options. Here are two proposed solutions that improve future ROI:
- Get operations expertise with your solution: First, instead of hiring in-house engineers and buying drilling equipment, work with a HaaS provider. Why? Because the provider of a comprehensive HaaS solution will not only be able to provide Hadoop expertise, but they will also be able to deliver a lower total cost of ownership and more predictable costs over time — factors that make for more reliable ROI calculations and better overall outcomes.
- Better predict benefits using a specific, proven case: An experienced HaaS provider will get an organization up and running with a proof of concept (POC) that addresses a well-defined issue, so that the benefits of a Big Data investment in production can be easily predicted.
Estimating Big Data ROI my feel like taking a leap of faith. However, if you’re choosing the right method of deployment and have access to operations expertise, it’s not a leap of faith at all. With these two factors in mind, predicting and getting future ROI on big data deployments becomes a purposeful stride toward increased analytical insight and improved business results.
Sign up for the free insideBIGDATA newsletter.