Big Data Saves T-Mobile Big Bucks

Moving to Hadoop-based storage and analytics platform from RainStor makes data analytics faster and less expensive.

June 25, 2014

4 Min Read
Big Data Saves T-Mobile Big Bucks

By moving to new Hadoop 2.0-based data storage from data store provider RainStor, T-Mobile has been able to dramatically increase the speed of its data queries, while sharply decreasing the physical size of the storage infrastructure.

As a result of the changes, the mobile operator can now more quickly identify and thus fix network and device problems, improving customer experience and more effectively targeting network upgrade dollars, says Art Popp, principal architect at T-Mobile US Inc.

RainStor Inc. formally announced its Archive Application availability for Hadoop 2.0 yesterday, and is targeting the telecom and banking sectors. RainStor's secret sauce is massive compression that allows much more efficient storage of raw data to use in conducting queries that help telecom operators such as T-Mobile assess where its customers are having problems. (See RainStor Launches Archive on Hadoop 2.0.)

"We store all the customer usage data and keep a log of everywhere you surf, and every destination from which you text something in a huge data store so we can run statistics on it and do very difficult queries, such as determining which handsets are underperforming in rural areas," Popp tells Light Reading.

Storing that volume of data has required a massive physical infrastructure: T-Mobile has been using IBM Corp. (NYSE: IBM) Netezza data warehouse appliances to store 15 billion rows of data a day on 30 racks of physical gear. (Relational database management systems measure stored data in rows). Using its new data storage platform based on Hadoop 2.0, an open-source software framework for Big Data storage and analytics, and the RainStor Archive App, T-Mobile will be able to store half a trillion rows of data at a time on eight racks of equipment.

After running massively parallel computing systems since 2006, T-Mobile was looking for less expensive ways to expand and found that Hadoop-based systems were less than half the cost of the IBM Netezzas, Popp says. The carrier deployed its first Hadoop infrastructure in April of 2013. But looking at the support needed for its LTE launch led Popp's crew to search for data compression options and that let to RainStor.

"We realized that with our current infra we would only be able to [store and analyze] 30-60 days of data and we wanted more if possible," Popp says. After a long proof of concept trial with RainStor, running what he calls the 30 hardest queries against seven billion rows of data, T-Mobile found RainStor ran "twice as fast in every circumstance and sometimes as much as 13 times faster."

All other features aside, Popp says, "a true 2X performance that saves us $2 million in gear -- that was an easy sell."

Being able to analyze customer data quickly saves T-Mobile money in bigger ways by enabling the operator to solve problems without just constantly adding expensive infrastructure to improve coverage, since some problems are not coverage-related.

Popp shared a couple of examples: T-Mobile was able to determine where the popular iPhone 5 handsets were underperforming Androids on its network based on customer data, and then conducted more intensive testing to determine the cause of the problem so it could be fixed. That process can help keep iPhone 5 fans from jumping ship.

His team was also able to identify an area of LTE coverage in which an abnormally high number of dual-mode phones were stuck on 3G.

"We know this annoys the customer and we know they will never call in and tell us," Popp says. "It's a subtle RF engineering thing we have to fix. But we may never know about it. Anytime analytics can reveal some subtlety that is beyond the noise of customer care tickets, that helps us."

Being able to use Big Data in this way translates to T-Mobile's bottom line in multiple ways: The first most obvious one is in customer retention, but the second one is in delaying network investment, even if it's just for a few months or a few quarters.

There was a time when the answer to most problems was to overbuild the infrastructure, Popp says, but those days are over. No matter how much capacity and spectrum is made available today, consumers will consume it quickly by finding new ways of using their smartphones.

As he notes, usage isn't likely to level out until all of the company's almost 50 million customers "can watch Netflix videos for six hours a day" on their mobile devices.

RainStor has specifically developed its compression system for stable data stores such as those from telecom and banking, which are machine-generated and don't change much, says Chief Architect Mark Cusack.

"What we are bringing to bear here is the specialized compression, security, query and data lifecycle management and governance capabilities specific to telco log data," he says.

— Carol Wilson, Editor-at-Large, Light Reading

Subscribe and receive the latest news from the industry.
Join 62,000+ members. Yes it's completely free.

You May Also Like