Loraine Lawson has an interesting post over at IT Business Edge that highlights a recent Forrester Research study on the Future of Software as a Service Technologies. In the study, Forrester predicts minimal success for BI and Integration delivered as SaaS.
This news is no surprise to me.
Back in late 2007, Goldman Sachs released "Getting SaaS savvy – successful investing in on-demand", a research report on SaaS from an investment perspective. (There’s a summary of the Goldman report at SandHill.com). Goldman used 10 criteria to assess applications and determine whether there are well suited for a SaaS model (and potential investment) . Among those, the amount of integration needed was a key criteria. BI and data warehousing fared poorly. Integration on demand was not included in the report, but using the same criteria, it would not have been done well either.
This news isn’t "bad news for SnapLogic" either. As a software vendor, we have always believed the customer should be able to control their deployment, so SnapLogic runs portably on multiple platforms, virtual machines, and Amazon Web Services.
There’s definitely ambiguity around the term ‘SaaS integration’. Integrating on-demand applications is not the same as on-demand integration; there are big differences between the two.
The decision on where to run the integration process should involve the location of the data, security constraints, and the customers preference.
If the bulk of the data is already on premises, then integrate by pulling the necessary SaaS data on-premises. Security wise, this approach doesn’t introduce major new issues related to the data.
If the bulk of the data is SaaS, then a hosted or on-demand approach to integration may be more appropriate. Security wise, it may not be possible to send on-premises data out, and that has to be taken into consideration as well.
One size does not fit all in data integration; there are too many changes afoot in the IT landscape, and everyone need to have options.