October 16, 2012

Mortar Takes Aim at Hadoop Usability

Ian Armas Foster

“Have a pile of under-utilized data? Want to use Hadoop but can’t spend weeks or months getting started?” According to fresh startup Mortar, these are questions that should appeal to potential Hadoop users, who are looking to wrap their arms around the elephant without hiring a phalanx of admins to oversee the operation.

Mortar claims to make Hadoop more accessible to the people most responsible for garnering insight from big data: data scientists and engineers. The young startup took flight when a couple of architects at Wireless Generation decided that big data tools and approaches were complex enough to warrant a new breed of offering–one that could take the hardware element out of Hadoop use.

Hadoop is a terrific open-source data tool that can process and perform analytics (sometimes predictive) on big data and large datasets. An unfortunate property of Hadoop is its difficult utility. Many companies looking to get into big data simply invest in Hadoop clusters without a vision as to how to use the cluster or without the resources, human on monetary, to execute said vision.

“Hadoop is an amazing technology but for most companies it was out of reach,” said Young in a presentation at the New York City Data Business Meetup in September.

To combat this, Mortar is building a web based product-as-a-service in which someone need simply need log on to the Mortar website and then they can start writing the code allowing their pile of data to do what it wants. “We wanted to make operation very easy,” said Young “because it’s very hard to hire people with Hadoop expertise and because Hadoop is sort of famously hard to operate.”

According to Young, Mortar’s ease of use comes partially from operating in Python, a favored language of engineers and data scientists. Further, users would be running their data on Mortar’s cloud-based Hadoop clusters. After the computation is complete, the data could be erased or left in the cluster. The latter option reduces the need to continuously open up and shut down Hadoop clusters, a not inexpensive process for large datasets.

“We built a platform-as-a-service to take care of operations and we focused on making Hadoop work with Python because data scientists and engineers often use Python to process data…When you log into Mortar, you can immediately write code into your browser without having to install anything.” From that processing, Young claims the data can be taken from the cluster into wherever the user wants it to go.

Mortar thinks its approach is unique. Whereas many vendors are going after business and end users with their Hadoop/BI tools, Mortar caters specifically to engineers and data scientists. Data scientists, especially those with Hadoop experience, are few and far between. Young says so himself, “If you’ve tried hiring data scientists, you know it’s really hard right now, and you’ve tried hiring data scientists who are also fluent in using Hadoop and big data tools, it’s even harder.”

The notion is that a company should get the most out of the data scientists they manage to hire. Initial results are promising, as Young claims that, through Mortar, data scientists were able to successfully use Hadoop in under an hour.

Put simply by their website, “Mortar is Hadoop in the cloud—an on-demand, wickedly scalable platform for big data.” At the very least, Young and Mortar are worth keeping an eye on from here on out.

Related Articles

Cascading into Hadoop’s Golden Era

Six Super-Scale Hadoop Deployments

How 8 Small Companies are Retooling Big Data