MemSQL Pushes HTAP Ball Forward
MemSQL used the second full day of the Strata + Hadoop World conference to launch a new version of its distributed SQL database that pushes forward its hybrid transactional/analytical processing (HTAP) strategy, which is gaining steam across the industry as a blended form of computing.
MemSQL is part of a new class of in-memory, horizontally scalable, relational databases that are gaining momentum for the capability to ingest and analyze large amounts of data in near real time. The database, which was first released in 2011, has been adopted by companies like Shutterstock, Pintrest, Comcast, and Ziff Davis that need to power operational applications with big data needs.
While the MemSQL database is primarily used to power operational applications, the software differentiates itself by adding analytical capabilities to the mix. It’s a hybrid form of database that research firm Gartner has dubbed HTAP and that Forrester calls “translyitical.”
The approach has uses across industries, says Gary Orenstein, MemSQL’s chief marketing officer. It’s used heavily in ecommerce settings, where companies want to know what products are being viewed or put in a shopping cart, and which ones are being transacted or abandoned. It’s used in the financial world, where fund managers are capturing financial data to build trading models or to build real-time portfolios. It’s used in the energy exploration business, where companies are tracking sensor data from heavy machinery for proactive maintenance scheduling.
“The common umbrella across everything is real time analysis,” Orenstein says. “Whether or not it include a heavy transactional component, they all include a desire to see what’s going with data in real time.”
With today’s launch of MemSQL version 5.0, the San Francisco-based company is boosting its analytical chops that should give customers faster results from analytical queries. Orenstein recently briefed Datanami on the new release, in particular the changes made to the MemSQL query optimizer and the query execution engine itself.
“What we noticed with customers doing ad hoc and exploratory analysis is that first query was not living up to their expectations,” Orenstein says. In an effort optimize the upfront query code generation, the company adopted Low Level Virtual Machine (LLVM) compilers, and as a result, the first queries are returned much more quickly than before.
“It gives us up to a 10x improvement in the performance of the first query, so the result is a more fluid experience for analysts doing ad hoc analysis or data exploration using Tableau (NYSE: DATA), Zoomdata, Looker, or MicroStrategy (NASDAQ: MSTR),” he says. “The net result is customer who in the past had first queries that were taking a moment longer than expected–that goes away.”
MemSQL CEO Eric Frenkiel will be at the Strata + Hadoop World conference this week to talk about version 5. Frienkiel will be on stage JR Cahill, a senior solutions architect at Kellogg, to talk about Kellogg’s use of MemSQL.
Cahill says that enterprises that adopt solutions to eliminate batch processing will overshadow those trapped by ETL. “At Kellogg, a primary mission in IT is to build a seamless platform delivering instant analytics to optimize cereal distribution nationwide,” Cahill says in a press release. “MemSQL 5 took us from a 24-hour process to one in under an hour. That time saved goes directly to the bottom line.”