Aerospike brings NoSQL to the core of financial services (Article mentioning Aerospike - April 21, 2016)


This article mentioning Aerospike ran in FierceFinanceIT and other publications. It is reprinted below.

April 21, 2016 | By Renee Caruthers

As financial companies increasingly leverage big data, they have explored and embraced non-relational databases for their most advanced analytics. But NoSQL database provider Aerospike envisions bigger use cases in finance.

“There is a huge amount of investment in those technologies for analytics uses, and it’s easier to go to market with, but the operational use case where you are really part of a business in a core way and every transaction flows through you, they are still using mainframes for that, so we thought that was a more interesting under-invested in area,” says Brian Bulkowski, Aerospike CTO and co-founder.

Aerospike designed its distributed in-memory database technology to support what it calls “speed at scale.” The six-year-old company got its start with ad tech companies that needed to serve and analyze ads moment-by-moment in an Internet environment across vast numbers of sites. For financial firms faced with merging the world of sophisticated analytics and massive speed demands with settlement systems operating from relational databases in mainframe systems, Aerospike aims to bridge that gap. Its early forays into the financial services world are in augmenting mainframe environments.

“You need a lot of your back office settlement systems for compliance reasons to still be in databases and traditional relational mainframe systems, and that’s not going away. But those things can’t take the read load if you are doing advanced risk calculations and if you are doing things like mobile retail brokerage loads. You can’t hit the mainframe every single time someone brings up their app.”

For many financial companies, the adoption of NoSQL databases, which do not use the traditional tabular structure of relational databases, means first overcoming the mindset of thinking in a relational database way.

“It’s always interesting interfacing with project teams because they will come and say ‘I have this table, and I want to do a query structured like this,’” Bulkowski says. “We say, ‘Let’s explore what you are really trying to do.’”

Bulkowski says relational databases are fit for purposes such as ledgers or bookkeeping applications that involve tables of transactions. NoSQL has been gaining ground in other analytic areas because it can be more efficient at accessing non-tabular data.

“When you are talking about tracking all the behavior of all the IP addresses on the Internet, and using that as part of a fraud calculation, relational isn’t buying you much, and it’s usually even getting in your way,” Bulkowski says.

Bulkowski says the company’s in-memory distributed technology allows it to deliver massively higher performance at a reasonable price. The company sees loads of millions of transactions per second with a small number of “garden variety” Intel Linux servers. An Intel test with Aerospike conducted a year ago reached 1 million transactions per second on a single server, and that was going to Flash storage, so it was persistent.

That’s where Aerospike envisions bridging the gap between traditional mainframe applications in financial firms and complex, fast-paced calculations. Bulkowski sees the technology sitting in front of or next to a mainframe in financial firms as part of a next generation operational system.

“How do I deal with the new demands of risk calculations to make sure that in a much more complicated and highly leverages world I’m actually obeying all the things that I need to obey? Those calculations are more sophisticated and trying to actually hit your system of record with that kind of load doesn’t work,” he says.