Couchbase replaces its storage engine with homegrown ForestDB

NoSQL vendor Couchbase has built a data store called ForestDB that will underpin future versions of the company’s database offerings, and the company says it’s faster than the Google-built LevelDB or the Facebook-built RocksDB. ForestDB is open source and currently in beta mode, which means it won’t be included in the 3.0 version of the company’s flagship Couchbase Server slated to be released next week at its annual user conference.

Couchbase has been focused on improving database capabilities for mobile applications for years, and ForestDB keeps up that tradition. According to Ravi Mayuram, Couchbase SVP of engineering, the technology was designed with the goal of maximizing both efficiency and performance on mobile devices and solid-state disks. In Couchbase’s tests, ForestDB was six times faster than LevelDB and RocksDB on both hard disk drives and solid-state drives, and was six times faster than SQLite on mobile devices, he said.

Mayuram also said ForestDB proved four times more efficient than either LevelDB or RocksDB. It was five times better than RocksDB and 20 times better than LevelDB with regard to write amplification — that is, the number of times a database has to write something to a solid-state drive before it’s actually stored. This is critical, he explained, because the less you tax the drive on each transaction, the longer it will last.

Some people question the value of storage efficiency because space is becoming so cheap, but, Mayuram said, “There is a logical end to that stuff.”

Source: Couchbase
Source: Couchbase

Despite the comparisons to alternative database engines, Mayuram said ForestDB isn’t really meant to be a competitive project or technology, except to the extent that Couchbase hopes ForestDB’s specs and features will help the company attract more users. If an open source community happens to grow up around it, the company will help out where it can.

Really, Couchbase is trying to establish a reputation as the innovation leader among NoSQL vendors, said CEO Bob Wiederhold. He points to the company’s N1QL query language as a major improvement on the data access side, and thinks ForestDB brings the same advantages on the performance side. The better that NoSQL databases perform, the easier it will be for companies to choose them over relational databases for certain tasks.

At that point, it becomes a matter of which NoSQL option to choose. “We think [ForestDB] is a very innovative technology that’s going to allow us to increase the leadership we have on performance against Mongo and DataStax,” Wiederhold said. “In the medium to long term,” he added, “[Couchbase’s pace of innovation is] going to serve us extremely well.”

He’s right about one thing, at least: A NoSQL space that started off with numerous projects and companies a few years ago has pretty much boiled down to a three-horse race between Couchbase, MongoDB and DataStax (with its commercial Cassandra offering), at least when it comes to securing those lucrative large-enterprise deployments. Each company has raised mountains of capital and each database has its own sweet spots and its own advantages. But assuming most companies will settle on one NoSQL option to complement their existing relational systems, every little improvement helps.

Feature image courtesy of Shutterstock user Andril Malkov.