ScaleOut Software is a company producing memory-based storage and analysis products. Today the company is announcing general availability of its hServer V2 product which allows MapReduce to be run on live data. The product has a self-contained execution engine for MapReduce which has tested out to increase Hadoop execution times by some 20x.
The initial product delivered low-latency data-access for Hadoop whereas this release goes to the next step and allows the actual execution of the analyses to occur in-memory. It allows for concurrent access of data set at the same time as MapReduce analyses are being performed. As such it will be of interest for use–cases where data is changing at a rapid pace and analyses need to be run quickly on this fast changing data.
Interestingly hServer doesn’t actually require Hadoop to be installed – rather the product integrated the MapReduce functionality and selected Hadoop components within the in-memory data grid – thus reducing installation times. It is compatible with most Java-based applications that have been developed for standard hadoop distros, thus further reducing the barriers to using Hadoop in-memory. All that is required is a one-line code change on existing applications
It is available in a free community edition that is limited to four servers and256GB of data sets – beyond this the commerical versions are licensed on an annual subscription.
This release will be attractive for industries that have both high data change rates and a need to product rapid analyses – financial services and industrial application spring to mind. In-memory isn’t for everything (and neither, for that matter, is Hadoop) but where timing is critical, the ScaleOut offering makes sense.
It’s also perhaps an indication that in-memory analytics is getting more attention – this offering obviously stands competitively alongside SAP‘s HANA and others, it will be interesting to see how this part of the market develops.