This site is operated by a business or businesses owned by Informa PLC and all copyright resides with them.Informa PLC's registered office is 5 Howick Place, London SW1P 1WG. Registered in England and Wales. Number 8860726.
WX2 version 7.0 improves in-memory analytics and reliability, cuts scheduled downtime.
With just over 30 customers and a two-year-old U.S. headquarters in Chicago, Kognitio sounds, at first blush, like another of those numerous data warehouse appliance startups. But in fact, the UK-based company has roots that go back more than 20 years to a company called Whitecross Systems. The firm was a pioneer, and it helped early customers like British Telecom query huge volumes of data long before most people talked about the "data explosion."
Way back in 2005 (anchient history, really), Kognitio had its own data warehouse appliance. But before entering the U.S. market in 2008, the company wisely decided to lose the proprietary hardware and focus on its WX2 in-memory database. Like many alternative databases, WX2 is designed for highly scalable, massively parallel processing -- now on commodity hardware -- but its differentiator is mature in-memory analysis and management capabilities.
Yesterday, Kognitio released version 7.0 of WX2, which the company uses as the foundation of highly scalable data warehouses appliances built on the blade server hardware of your choice (with HP, IBM and Sun being the leading choices). Upgrades in the new version include improvements on WX2's in-memory analysis capabilities, better fault recovery and the elimination of an old Achilles' heel for WX2 when it came to scheduled downtime.
In-memory analysis is a differentiator for WX2. The product's database optimizer is adept at RAM memory management, exploiting two core techniques. "When you run a query, WX2 will pull data from disk, put it into RAM and do the analytics there, or you can 'pin' data in memory," explains John Thompson, CEO of Kognitio's U.S. operations. "You can pin an entire database schema, fragments of tables, vertical slices of tables, horizontal slices of tables or time-dependent slices into memory. Any way that you can process a view in a relational database can be pinned into memory in WX2."
WX2 deployments can include as much memory as customers want -- sometimes upwards of tens of terabytes of RAM. The key benefit, of course, is fast analysis, and with pinning and memory management, the system can be optimized for mixed workloads and hundreds of concurrent users.
"WX2 demonstrates absolutely jaw-dropping performance," states customer Fiachra Woodman, IT Director at Loyalty Management Group. "At any one time, more than one hundred users from our user community query data in WX2 at the same time and there is no degradation to analytical performance."
Large-scale in-memory system optimizations performed for the 7.0 upgrade are said to have helped WX2 architects improve inter-node communications to better support system scalability. Other key upgrades in version 7.0 are aimed at reducing downtime. For example, automated fault recovery features were bolstered to ensure continuous operation.
"We've always had fault recovery, but we've improved it so that if a disk fails, a network connection dies or entire server fails for some reason, WX2 continues running and recovers in an automated matter," Thompson says.
To reduce planned downtime, WX2 version 7.0 eliminates an old drawback tied to the product's linear file system. Since this design adds new data to the end of the file system, at some point you have to reorganize the database to reclaim space from deletions and other changes. That used to mean taking WX2 offline, but a new feature called Reclaim introduced in WX2 7.0 continuously handles that task as a background process, thereby eliminating scheduled downtime. The latest version also extends multilingual support to include Asian languages.
Kognition deployments are "about a quarter of the price of Teradata and half the price of Netezza when you compare apples to apples," Thompson says. Entry-level deployments, including hardware, are said to typically be in the $250,000 to $300,000 range. The company also has a cloud-based, data-as-a-service offering with pay-as-you-go pricing.
We welcome your comments on this topic on our social media channels, or [contact us directly] with questions about the site.