To suggest that enterprise application integration (EAI) might be nearing the end of the road is both contentious and provocative, writes Bloor analyst Philip Howard. Do we really mean it? Well, yes and no. No, we don't mean that EAI as an approach will disappear or that it does not fulfil a useful function. But yes, we do expect it to be merged into another technology - data federation - and to be subsumed by it.
Data federation offers a superset of the facilities offered by EII (enterprise information integration) in that it offers, or may offer, update as well as query capabilities across multiple data sources. We also expect that data federation will come to offer a superset of the facilities offered by EAI.
Consider the major distinguishing feature that separates EAI and data federation. It is not that one is application to application and the other is application to data source. After all, a data federation solution can pass a result set to a message queue as easily it can to an application. Nor is it, as we have previously discussed, that one is a push technology (EAI) and data federation is a pull technology. This is because, actually, data federation products often include data replication capability. For example, IBM's DB2 Information Integrator includes both replication and federation capabilities. Even without replication, change data capture is a commonly used push mechanism implemented with data federation products.
No, the real difference between EAI and data federation is that the former is transaction-oriented while the latter is not. Or, more specifically, EAI is record based and data federation is set based.
But - and here is our point - a record is simply a very small set. In other words it is a simple extension for a data federation vendor to extend its capabilities to support EAI. Conversely, to extend an EAI solution to provide set-based processing would require a major re-architecting of the product. Thus any movement can only be one way.
Of course, this isn't the end of it. There are a variety of facilities that you would expect to see in an EAI environment that will not necessarily be in any current data federation product. But none of these features amount to rocket science and, in any case, data federation is a much newer discipline than EAI and we would expect it to take a little while to catch up, let alone overtake its bigger brother.
So, is the demise of EAI realistic? To the extent that there will soon be products that do both functions under a single umbrella - yes. We recently spoke to one vendor who plans to announce exactly this strategy later in the spring, and we know of several others that are moving in this direction even if they cannot yet be explicit about their product plans.