Apache Hadoop users will soon be able to analyze data as it is streamed from its source, thanks to a partnership between data-warehouse software provider Informatica and Hadoop distributor MapR.
The companies are integrating their products so that the new world of big data analysis can work more easily with more traditional data warehouse implementations.
[ Also read the InfoWorld story Supersize me: Hadoop upgrade will handle even bigger data. | Also on InfoWorld: Look before you leap into Hadoop. | Explore the current trends and solutions in BI with InfoWorld's interactive Business Intelligence iGuide. | Keep up with the latest approaches to managing information overload and staying compliant in InfoWorld's Enterprise Data Explosion newsletter. ]
Specifically, the companies are writing a connector that will ingest data streamed from Informatica's Ultra Messaging application into a MapR Hadoop implementation.
Ultra Messaging copies log file entries, transaction data and other forms of high-volume, continually updated content onto a messaging bus, so it can be reused and analyzed by other systems. Hadoop is a data processing platform, one that can be used to store and analyze large amounts of data in varying formats.
One disadvantage to Hadoop is that it is designed for batch processing, explained Jack Norris, MapR vice president of marketing. With the standard edition of Hadoop, the underlying file system, HDFS, requires that a data file be closed before it can be analyzed. This can be problematic when trying to analyze a flow of constantly updated data. The administrator must pick arbitrary times to close the file for analysis. As a result, "You are knowingly dealing in old data," Norris said.
MapR's distribution, however, is unique in that it allows data to be read even while the file the data resides in is still open and being written to. By connecting MapR with Ultra Messaging, the combined system will offer the ability to analyze data in near-real time as it comes off the message bus.
With Hadoop, users can then combine this live data with other types of data, providing a wider breadth of data to analyze. "With Hadoop, [analysis] is not just done on a single data source. It's the combination of different data sources," Norris said.
This combination of technologies would be handy for time-sensitive pattern recognition tasks, Norris said. One such task is fraud detection, in which a financial institution would need to spot the misuse of its credit cards as early as possible. While computer systems have long been used for fraud detection, using Hadoop in conjunction with a stream of live data provides more data sources to monitor, along with the ability to identify infractions more quickly. "You can look across an entire portfolio of transactions and detect small frauds earlier," Norris said.
At least one other technology has been created to tackle the problem of real-time big data analysis. Last year, Twitter purchased BackType, and subsequently released as open source the company's Storm stream data analysis software. Twitter itself uses the software to spot emerging trends from its users.
In addition to Ultra Messaging, the two companies are building connectors to other Informatica data warehousing tools, including bidirectional connectivity with Informatica's flagship PowerCenter and PowerExchange data warehouse applications. MapR data will be able to be backed up in Informatica Data Replication and Informatica FastClone. Also, the community edition of Informatica's HParser, a Hadoop file parser, will be bundled with the MapR distribution.