We have recently Bloor Research web site. While I don't usually blow my own trumpet about our reports this has some interesting conclusions that are worth sharing.
We surveyed more than 200 companies to find out how they were using data integration platforms and what were the associated costs over an extended period of time and, arguably more importantly, how effective those costs were when spread across the number of projects that the platform supported.
While I know that Pervasive Software coming out on top in this last category is probably the headline news coming out of this research, I would like to highlight a few of the other points that we discovered.
The first of these is with respect to what data integration is and is not being used for. While there was some variation by tool (for example, Informatica PowerCenter was used most often for traditional ETL) we found that, across the board, the most common use of data integration platforms was for data migration and conversion projects. This is perhaps a reflection of the fact that you only have one enterprise data warehouse and loading it is considered a single job in perpetuity but it is an interesting result nevertheless.
Conversely, of the five use cases we put to users, using data integration platforms to support SOA initiatives, was the least popular choice by some margin. Probably this reflects the fact that many companies adopting SOA are still very focused on the application side of that equation and have yet to wake up to the importance of the data half. I therefore expect this usage type to grow significantly.
A third interesting fact emerged when we looked at the length (and complexity) of the projects for which different tools were used. We broke these down into short, medium and longer projects and the general trend was for there to be an increasing number of projects as the length of the projects increased. However, this was not the case with Microsoft or open source solutions which had peak numbers of projects for medium length, suggesting that these are not really seen as suitable (or not so suitable) for the most complex projects. At the other end of the spectrum, vendors like IBM and Informatica had a definite skew towards longer and away from shorter projects. Of suppliers that fit the general pattern Pervasive had the most even spread of project support.
Finally, open source solutions did not fare as well as I had expected. Of course they were inexpensive but they were not reused across as many projects as the proprietary solutions, which meant that their total cost of ownership per project went up. This may be because it is relatively early days for open source companies in this market and, as products mature, then we will see an increase in the number of projects but that is not the case at present.