Using SOA to cut out the middle man

Bookmark Blog Post

	del.icio.us
	Digg
	DZone
	Furl
	Reddit

Hussein Badakhchani's Blog | June 30, 2006  12:19 AM | Comments (0)

I came across a good article in International Banking Systems titled "Cutting out the middleman". It described how Merrill Lynch managed to cut costs without ditching its maninframe systems using SOA. Nothing particularly interesting about that although I should add that many organisations decide to ditch their mainframes because they come to the conclusion that technology renewal is ultimately cheeper than wrapping mainframes in SOA (this is certainly the case were I am currently working), so Merrill Lynch deciding to keep its mainframes is noteworthy.

The interesting part of the article was the way they implemented SOA. It seems they decided "the owner of the service should be the one to publish it" and the article stated that up till now the thinking has been that publishing of services is a distributed problem and thus should be handled by Java programmers. Instead Merril Lynch empowered their mainframe programmers by giving them the tools they needed to publish their services.

Jim Crew who lead the project at Merril Lynch (now at Software after it acquired the team and technology) is said to be "skeptical of solutions that involve creating a middle tier" as he argues "that it is one more black box which often becomes the source of failure". I have to say that this rings true to me. Jim continues to moan about not having enough visability of transactions and how transactions run into the middle tier and just stop there. Well any middle tier is only as good as the developers and requirements used to implement it. I see solutions that meet the business requirments 100% and have the signed off documents to prove it. The problem is no one gave much thought to supporting and debuging the damn solution once it goes live, and hence the business finds it's wonderful solution becomming just another problem.

The article goes on to list some of the requirements they used to implement their tool (now called SOLA - Service Orientated Legacy Architecture) to publish their mainframe applications as webserivces - reliable, offer performance monitoring, good debugging, execellent security, and service level agreement. I can't remember the last time I read "good debugging" in a requirements document, I guess if their middle tier was developed with the same requirments they may have been less inclined to see it as the problem it was.

Comments