≡ Menu

First Look – Kognitio WX2


The folks at Kognitio gave me a quick overview of their WX2 product recently. WX2 is aimed at three distinct areas:

  • Software platform for high performance analytics
  • Data as a Service – it is being offered hosted to SMBs for instance
  • Data Warehouse Appliances

WX2 came out of the UK and is the end result of some 20 years of development. It is a full functional relational database designed from the ground up to have linear scalability on parallel hardware – everything runs in parallel – and very high performance for high user concurrency, complex queries and mixed workloads. It uses x86 servers as a platform and is easy to reconfigure, resilient and highly available. They claim very easy and rapid installs – hours rather than days – and that supporting even very complex reports only takes a week or two of setup. The product requires minimal DBA support and is designed to support incremental design – the constant addition and refinement of tables, columns and queries. The use of generic hardware makes it cheaper to maintain and they focus on ensuring that staff spend their time on data analysis not on understanding or preparation. Plug-ins are supported for extensibility (you could run an executable model as a java plugin for instance) and these run in parallel too.

The demo of some queries was very impressive with even complex queries such as counts across calculated attributes, self-joins on huge (17B row) tables and intensely analytical queries running very fast. They have some partnerships with analytic modeling vendors, such as KXEN, and their sweet spot is where good performance is required for unprecitable queries, growing and complex data and complex queries. They compete with data warehouse products.

The technology clearly offers the potential for closing the gap between operational and analytical processing by supporting both, although it sounds like most of their customers are still stuck in the past and use it only as a high performance data warehouse. Some are moving in this direction with large amounts of operational data (such as Call Data Records in telecom) being loaded and queries directly without any aggregation.