We heard a lot of talk a few months ago about about the death of service oriented architecture (SOA). This was mainly by the kind of pundits who had over-hyped it in the first place. Since then, we have seen various rebuttals to the claims that were made, but a degree of uncertainty about the status and role of SOA persists. So …
...is unpopular with those who talk a good fight but would not know a basic profile if one hit them in the face. Their stock in trade is procrastination as it eats money with no (percieved) risk.
Why do I feel I should have mixed some capitals in there in aManFromMars styley?
SOA = Enterprise Architecture = bulls**t.
SOA, the whole purpose of which is to generate consultancy revenue.
Of course some people will try to defend it, they have a vested interest.
Haven't you heard?
It's SAAS and Cloud now.
SOA just another IT re-badge
SOA isn't really anything wonderful, it is just an extension of the IT industries continuous progress towards cost reduction, modularisation and re-use, at the expense of performance. No good programmer ever writes code that can't be re-used somewhere else, I still use code I wrote 20 years ago, because it does what is supposed to do. I can wrap that code in any interface mechanism that I want to.
I now design large systems, and I am consistently amazed at client, and internal system guru's who thing that everything should be SOA, and then wonder why it won't perform. In one project a Solutions Architect seems to think that the system has to have every module exposed as a service via an XML UDDI definition, despite having been told umpteen times that it is only sensible to expose external business services and not internal technical services. He also doesn't seem to understand that ESBs don't actually need a web service interface to communicate with applications, having fallen into the trap of believing that SOA = Web Services, and failing to understand that just because something can be a service doesn't mean it should be. Sadly his view is reflected by far too many sales oriented consultants, who don't have to deliver the solution. SOA also suffers from what Martin Butler described as the "Me Too" culture that pervades the IT industry, and business too for that matter, "monkey see, monkey do". SOA is a great Architecture Model, it isn't a tool, and it isn't Web Services and ESBs. SOA builds bridges, and you wouldn't use bridge architecture to design a house would you.
gimmick of the week
SaaS and the cloud -- the backlash is due in about six months from now, by my reckoning. Don't worry, though, there'll be some exciting shiny new game-changing paradigm shift (in the buzzwords used in press releases, at least) not long after that.
I got into professional IT in the mid 90s; the buzzwords-du-jour were OLAP, datacubes, data warehouses, RAD and MIS. (oh and layer-3 switching, heh!) Since then many new fads and crazes have come and gone, the only change for me is how seriously I take the trade rags... ie., "not at all".
SOA is Web :)
It never changes now, Web is it, sure you could use your own protocols, but why bother, web means you can turn it on anywhere and access anywhere, double edge but flexible.
Cloud smoud, distributed virtualisation is all it is, and it is not that great, OpenVS and Jails are better for real world virtualisation, it is only if you really need the distributed in a hurry, and still Erlang is wiser.
Very small niche areas, but web is huge, and distributed will be massive but nodes and state machines are what you want not virtual single machines. It is a big paradigm shift.
Distributed virtualisation costs too much in overhead, and won't be used that often, and so it will be a miracle when it does work.
Still room for it, Xen will lead it, but nothing more than another way of doing things, it is not at core the large shift, distributed and concurrent is.
No I don't
seems to me like SOA is a bit like going to the gym
you're sold by the pictures of attractive, happy, sculpted men and women, and agree to sign a few dollars a month away
then you start realising it's not just about lifting weights and going on the treadmill, they're asking you to change your diet too
after a few months or so, you may not look quite like the pictures you were sold, but you are more awake inna mornings and your partner likes the way you're a bit more toned and you're eating a bit better
is that good enough? sometimes
break out the bubbly for another Nobel prize
SOA; where practice regularly disproves theory, and reduce it all to lowest common denominator. A bit like ATM will bury Ethernet.
Analyst firms/consulting houses alway chance their arm at grandstanding a concept, regardless if its appropriate or not - shoo'em out the door: SOA has merit, *where* it's appropriate: why the hell would I change my dedicated app:app hi-perf with 'get-in-line' on a ESB?