VoltDB Turns to Real-Time Analytics with NewSQL Database
NoSQL databases and their NewSQL cousins are typically thought of as operational data stores. They’re the fast and flexible databases you bring in after your transactional application has maxed out the traditional relational database powering it. But for the folks at VoltDB, the latest 4.0 release of its NewSQL database is all about enabling real-time analytics.
VoltDB is an in-memory NewSQL database that customers generally use to power data-hungry Web and mobile applications that exceed the capabilities of standard relational technology. The database talks JSON and SQL fluently, includes the sharding and horizontal clustering capabilities that people traditionally associate with NoSQL databases, and is ACID-compliant (it manages durability by writing data to disk through a combination of snapshots and write-ahead logging).
According to VoltDB CTO Ryan Betts, customers are requesting more analytic capabilities in the database so they can power real-time decision making. In many cases, they can’t get data into their existing online analytical processing (OLAP) or data warehouses fast enough. So VoltDB took it upon itself to enable more analytical processing within its database.
“People are looking at OLAP systems and in the cases of specific analytical cases, those OLAP systems aren’t fast enough,” Betts tells Datanami. “So they’ve begun looking to in-memory systems to speed up specific dash boarding analytics, simple grouping, and sometimes ad-hoc filtering of columns for simple reporting and real-time decision making.”
On a technical level, the company delivered this capability by lifting the cap on the number of read-intensive queries the database can process. Those types of queries can 10 to 20x faster in the version 4.0 release of the product. The company also bolstered its support for SQL and enabled “broader set of queries” to be presented to the database, Betts says.
It’s all about enabling quick decision-making when surrounded by a sea of fast moving data. “The analytics we’re talking about are the analytics that need to be done in a moment–outside of a batch process, in real time–in order to reach a decision about how do I respond to the user,” Betts says.
One of VoltDB’s customers is a large bank that wants to make offers customer offers in real time as they interact through the Web or mobile channels. The product is also used by telcos and others to manage network traffic. “They have gobs of data that’s latency sensitive, and managing and understanding latencies is critical for managing quality of service,” Betts says.
Moving forward, VoltDB plans on improving the core transaction processing capabilities of the database, as well as the analytics. The company currently has some customers running distributed clusters of VoltDB servers running on upwards of 10TB of RAM. That’s enough memory to power millions of transactions per second, and make a few basic analytic decisions at the same time.
In other news, this week VoltDB also announced that it saw a 298 percent increase in revenue in 2013, and a 368 percent boost in bookings, as compared to 2012. The company has more than 325 paying customers. The company isn’t at the top of the NoSQL heap, but it will be helping to make 2014 an interesting year in the space.