At TDWI in Las Vegas this week, Vertica announced version 4.0 of its Analytic Database columnar MPP technology. There are several main areas of improved functionality. First is a major revamp of the optimiser. As well as an improved cost model, this closes certain gaps, such as proper support for full outer joins. Next is significantly improved resource and workload management. In particular it is now possible to assign priorities to certain users or processes, and to assign run-time resources to processes (for example, to stop a large batch job hogging the system). There is also improved partition management, for example the ability to drop a partition while conducting a load.

Another area of improvement is in support for analytic functions. Version 4.0 brings time-series analysis, moving average and cumulative functions, which are relevant to certain types of use cases, such as securities analysis. Vertica supports MapReduce, but not as an in-database function. There is also Unicode (UTF-8) support, which opens up the possibility for Vertica of selling into more international markets (at present its sales are around 80% US and the rest in Europe, principally in the UK), for example Asian countries which require double-byte character support.

Additionally this release brings dramatically improved performance on large deletes and updates, which have traditionally been the Achilles Heel of columnar approaches. It may seem as if this should hardly be an issue for data warehousing applications, which are essentially read-only, but in fact many real-life applications do require data to be updated regularly, even if they are not OLTP applications. Finally there are improvements to the user interface of both the automated database designer and the administration tool, bringing in better graphical representations of memory and CPU usage.

Vertica continues to show strong commercial progress, with a doubling of its customer base in 2009, and over 120 commercial customers now using its technology. Customers include Verizon, JP Morgan, Experian and Sprint. This release demonstrates its continued investment in the core product, and its responsiveness to its customer feedback e.g. the incorporation of time series and moving averages, which are heavily used in the capital markets. Given the well-documented struggles of some data warehouse appliance vendors, it is encouraging to see Vertica continuing to grow rapidly. This also validates the columnar MPP approach to analytic applications as an alternative to traditional row-oriented MPP approaches.