There's nothing that new about in-memory databases, but with advances in hardware behind them, their use within low-latency applications is drawing increasing interest. Recent news from McObject and MemSQL underscores the trend.
McObject has released the eXtremeDB Financial Edition, which adds some financial markets functionality to its existing small footprint in-memory database.
The key additions for the FE release are support for columnar data structures, for storing time series data (making efficient use of the L1/L2 cache for each CPU), and a set of vector-based statistical functions pertinent to processing time series. These functions include logical and arithmetic operators, averages and variances, standard deviation, and user defined.
Also new in eXtremeDB FE is the ability to keep certain tables on a local server in a clustered environment for faster searching, especially where the table contains data that is common to a number of data structures, such as stock symbol. There is also a new performance monitor to observe how changes to the database impact transaction throughput, allowing fine tuning to reduce latency.
Meanwhile, MemSQL's news was its emergence from stealth mode to launch its eponymous in-memory database, together with details of a number of investors that are backing it to the tune of $5 million.
A key feature of its database is the translation of queries from SQL to C++ in order to boost performance. This allows it to offer predictable sub-millisecond response times for massive datasets, running on commodity hardware.
Behind MemSQL - founded last year by Eric Frenkiel, CEO, and Nikita Shamgunov, CTO - are a combination of venture capitalists and angels, including First Round Capital, IA Ventures, NEA, SV Angel, Y Combinator, Paul Buchheit, Ashton Kutcher, Max Levchin and Aaron Levie.