Managed hosting provider Sungard is getting into the big data space with a new Hadoop service that gives users on-demand access to the popular data-storage and processing platform. Called Unified Analytics Service, Sungard’s new offering joins the growing ranks of cloud-based Hadoop offerings from Amazon Web Services, (s amzn) Microsoft (s msft) and IBM, (s ibm)as well as startups such as InfoChimps and Mortar Data.
Running Hadoop in the cloud is gaining popularity as an option because although the open source software is a powerful tool for analyzing large amounts of unstructured data, it can be very difficult to build and maintain Hadoop systems. Using a managed Hadoop service means IT departments can get the benefits of Hadoop for analytics without having to know, for example, which version to use or how to design the system for high availability.
Sungard CTO Indu Kodukula told me during an interview that Sungard is really embracing the cloud aspects of the new service, both technologically and business-wise. On the business front, explained that customers will be billed monthly based on how many terabytes they’re using, and that the service is designed to make it easy for customers to import and export their data.
Technogically, Kodukula explained, Sungard is working with the University of Texas to make Hadoop run optimally in a cloud environment. While it performs just fine now, the new research — which Kodukula hopes will be done in 2013 — will let Sungard optimize its Hadoop service to run atop a multitenant infrastructure and to boost reliability and performance by working around the problematic NameNode architecture of the Hadoop Distributed File System.
However, the one issue nobody has figured out when it comes to running Hadoop in the cloud is the difficulty of uploading potentially dozens of terabytes into the new environment. Kodukula acknowledges that simply sending it a hard drive might be the easiest option, although customers also can choose to use the public Internet or a dedicated connection. Once the data’s in the cloud, though, he said it’s “absolutely reasonable” to deal with change to the data set over the network, and many customers will probably want to add new data straight from web sources or public data sets that don’t require such large transfers.
Sungard’s Hadoop service is in technical preview mode and will be generally available in Q3. Kodukula said early interest has been strongest among financial services, government and health care users.