ParAccel Database Update to Focus on Analytics, Performance

ParAccel announces plans for Version 2.5 of the ParAccel database, which will be released later in 2010 with a number of enhancements to boost performance.

Data warehousing vendor ParAccel is putting an emphasis on high-performance analytics in the upcoming version of its database appliance.

According to the company, when it releases ParAccel Database Version 2.5 in the second quarter of 2010, it will be armed with a number of enhancements designed to improve performance. For starters, ParAccel's advanced query optimizer, Omne, has been enhanced to "allow automatic decorrelation of ... broader ranges of embedded subqueries to further simplify and accelerate the development of analytic applications," ParAccel said Feb. 21.

"First it eliminates some barriers that as you're developing analytic apps you always have to be cognizant of, barriers essentially around query complexity," explained Rick Glick, ParAccel vice president of Technology and Architecture. "For example, we can handle skip level decorrelation query patterns, which is I believe unique to us. Secondly, by avoiding bad plans and always having consistently great performance we dramatically reduce the amount of physical design work and modeling that an analytic app developer goes through."

ParAccel also has focused on taking advantage of solid-state storage for the release, and is "partnering with Fusion-io ... to run PADB 2.5 on servers utilizing Fusion-io drives," the company said. "The combination of Fusion-io flash technology and PADB's advanced analytic processing architecture allows customers to do far more with fewer servers."

"One Flash card offers the same scan performance as a bunch of drives, and you can put multiple Flash drives in these [appliances]," Glick said.

Other performance boosts come under the hood, such as the appliance's handling of metadata in column block headers.

"If information in the block headers tells us that the values we are looking for can't be in that data block we never read the block from disk," Glick said. "On highly restrictive scan operations this yields a dramatic performance improvement."

In addition, the company has added column-level encryption to support compliance with the Payment Card Industry Data Security Standard. Barry Zane, CTO and founder of ParAccel, said in statement that the upcoming release builds on the database's "massively parallel, columnar design" to "shatter conventional limits on analytic complexity ... The result is greater decision confidence and powerful leaps forward in ability to execute for companies across a wide range of industries."