Business Information

Technology insights for the data-driven enterprise

Sergey Nivens - Fotolia

Tax software combines public and private cloud architecture

Intuit sees private and public cloud as a dynamic analytics duo that can quickly scale systems, control computing power and test planned changes to its website.

For security and privacy reasons, Intuit Inc. initially set up a private cloud architecture for analyzing the data -- much of it sensitive personal information -- that the company collects from users of its accounting and tax preparation software for individuals and small businesses. But it's now also tapping the Amazon Web Services (AWS) public cloud to increase its IT agility for some processing and analytics uses.

Being able to quickly scale systems up and down to apply the right amount of computing power to jobs -- and control costs in the process -- is what drew Intuit to the AWS cloud, said Rekha Joshi, a staff engineer at the Mountain View, Calif., company. For example, Intuit went to AWS to deploy a combination of the Spark processing engine and the DataStax Enterprise version of Cassandra, an open source NoSQL database management system. Currently, the duo is primarily being used to run A/B tests on planned changes to Intuit's website, with Spark serving as the analytics engine and Cassandra as the data store.

But Joshi said the long-term goal is to make the testing platform available to all areas of the business at Intuit. As a result, more users are likely coming, which means the platform needs to be able to scale in a flexible way -- something that she's confident AWS will be able to handle.

Advancements that cloud services vendors have made in data security protection in recent years have also made public cloud systems more palatable to Intuit for tasks such as A/B testing. "The hesitation was there," she said. "But, in fact, internal servers are not as secure as the AWS cloud."

The testing team at Intuit is using the core Spark engine along with the technology's Spark Streaming module for processing streams of incoming data. Joshi said the team is also beginning to explore how MLlib, Spark's library of machine learning algorithms, might aid in the testing efforts, but nothing has reached the production stage on that yet.

Article 3 of 9

Next Steps

The debate over public and private cloud architecture

Considering which data to store in the cloud

Many gateway options for the cloud

Dig Deeper on Big data analytics

Get More Business Information

Access to all of our back issues View All