The Protagonist
LinkSmart
This startuplocated in Boulder, Coloradoprovides an advanced analytics service to its digital publishing clients. LinkSmarts unique, in-content linking technology is used to optimize audience engagement and increase viewing periods. The analytics solution offered by this company enables digital publishers to employ evaluative metrics and link-management tools to direct audiences to the suitable content through fitting cross linking.
LinkSmarts involved analytics help publishers understand which content pages require additional links and how these links should be redirected to new destinations to increase viewing time.
The Problem
Operating Costs
As a big data startup with limited operating budget, LinkSmart thrives on cost savings. The analytics service they provide involves every word in every article on a site. The requirement of storing vast amounts of data and analyzing it for word-level parsing was something that relational databases could not handle. So, this company was looking for an analytics infrastructure that could churn big data multiple times to prove the validity of their algorithms.
As LinkSmart CTO Manny Puentes describes, When youve adopted Big Data at the core of your business, you have to continually analyze data and everything backs into the dollar. Every time the CPU ticks, dollars are being spent. Each decision we make is based on how to do more with less money.
The Solution
Hadoop with MapR
After much research into suppliers, LinkSmart discovered a solution that provided huge cost savings, while preserving simultaneous high-volume, high-speed, unstructured data analytics capability. They evaluated Cloudera Hadoop distribution with MapR, and found that this was the most scalable and the most fault-tolerant system.
LinkSmart determined that Hadoop with MapR offered an affordable but reliableenterprise-level, data storage and analytics platform with support for mission-critical production cycles. MapRs dependability coupled with Hadoops superfast processing speed and NoSQLs unstructured data support provided just the right big data environment for this startup business. LinkSmarts MapR Hadoop distribution runs on 10 Intel Xeon processors, equipped with 64GB memory and four 2-terabyte SAS drives with Ubuntu 12.04.
LinksSmarts Solution
Derived benefits
Ease of deployment and management: MapR facilitates NFS mount, which enables direct access to the Hadoop cluster via standard Linux commands. This capability greatly reduces system-administration headaches. LinkSmart, by reducing systems administration overloads, has freed up developers to concentrate more on their product rather than on maintaining the analytics infrastructure.
High-end performance: MapR facilitate superfast compilation of high volume data in a matter of minutes. The real-time processing capability yield just-in-time decision making for customers. As Puentes points out, the data processing that took 26 hours to complete before was now completed in 15 minutes by MapR.
LinkSmarts Puentes adds that This makes a big impact for our publishersgoing to real-time hourly reporting vs. daily reporting. They are able to dynamically add links and see the results in real time. This allows them to take advantage of breaking stories to affect reader behavior by the afternoon.
Cost savings: Before implementing Hadoop with MapR, this company was spending thousands of dollars per month on data processing. With MapR, these processing costs have been reduced by over 40%.
MapRs product support team: MapRs enterprise-grade support team matches up with the unmatched technology of the product.