Summer seems to be a time for reading through that huge pile of interesting articles and magazines that you can only find time to look at on vacation.
On flicking through my own mountain of stuff, I came across the Q2 edition of Financial-i, a magazine targeted at the Financial Services industry. One point to jump out at me was a comment from Paul Joynt of Nordea, a Scandinavian finance house. Paul was pointing out that SOA does not necessarily solve the problems associated with legacy integration.
The article, ‘SOA – is it worth the effort’, is available from the Financial-i site if you register, but Paul comments that covering a legacy system with a wrapper “so it looks like what you want” still leaves problems with the next level of change, because “it’s only a veneer”.
I think Paul has hit on an important point here. Different vendors in the SOA space have different approaches to addressing the problem of integrating legacy systems. Some will simply ‘hand off’ the request for legacy information to a tool from the legacy supplier – in the case of IBM mainframes this might mean using WebSphereMQ as the bridge, for instance. Others might approach the problem in some sort of screen-scraping or other interface simulation approach, where the legacy application is fooled into thinking it is running in its normal mode of operations. Yet more may generate code-based wrappers for each individual need, to be executed whenever a particular service is required.
To me, this all sounds too much like veneer in Paul’s terms. Although this might address immediate needs, future changes will continue to generate substantial additional work and the generation of more and more ‘special-case’ code and wrappers.
Instead, the best of breed legacy integration solution should embrace SOA and integration rather than try to fool it with wrappers designed to seal off the legacy world from the outside. Legacy integration should be about making the legacy system a full and active participant in the service definition and execution. For example, orchestration should be possible both outside and within the legacy environment. Services should be built with full participation from both sides. By taking this approach, the best of breed legacy integration tool will ensure that future changes will become easier, quicker, cheaper and more reliable.
For more information on the whole subject of legacy integration, specifically in the case of mainframe systems, Lustratus offers a free paper on the subject.