Monday, October 26, 2009

Bashing EAI

A bit of EAI bashing going on these days. I suspect a lot of people, myself included, who have successfully implemented an EAI style of architecture are probably scratching their heads going hmm what did I do so wrong? Answer: nothing.

SOA is the latest new kid on the block (actually Cloud Computing is but we are not going there) so the natural tendency is to bash everything that came before it. The practice of EAI had and does contain a ton of good architectural principles. Those are being thrown out and replaced with the notion that EAI and it's evolution the ESB style of architecture are all about products and not architecture. It's simply not true.

Don't get me wrong, SOA is a very good thing. My issue is the stark contrast being drawn between it and EAI. Yes there are differences but they share more in common that is being explained by the bashers. Abstraction, loose coupling, common data types, composibility, discovery and yes even standards all existed in the EAI world.

EAI, ESB's, SOA and even Cloud Computing all have their place in the modern enterprise. Understand your architecture which means what works and what doesn't, evolve where needed but don't throw your successful practices out with the bath water because of the latest buzz.

S0A-Manifesto

A group of very smart and respected folks have put together a SOA-Manifesto.  You can find the Manifesto - http://www.soa-manifesto.com/ .  Overall it's pretty high-level stuff but good stuff.  I have one big issue however with one of the principles.  "Intrinsic interoperability over custom integration".  If just glanced at it sounds like something that makes sense but the devil is in the details.
 
Intrinsic means in this context that services using some sort of standards should do the talking between your existing applications versus doing "custom integration".  This concept is going to be confusing to an average IT shop.  The question will arise what's the difference between putting in a service and putting in a custom integration solution?  How do I get to the last mile of the interface with a third party application?  Or do you mean my services should contain all the business logic?  But isn't that custom because you had to write it?  Standards based stuff is fine but it's still custom if you have to write the service.  Sounds like a lot of academic semantics I know but I think its important to understand some important principles.
 
Unless your entire business portfolio is going to be made up of custom services that you write then you are going to have to do some integration.  It's just the nature of most businesses.  My problem with "Intrinsic interoperability" is that I believe it will lead to a spaghetti set of custom integrations.  The very thing it was trying to avoid.  Custom is custom.  In other words if you didn't buy it then it's custom.  Whether it's "standards based" or not has no bearing on whether it's custom or not or whether it becomes spaghetti or not.  I think well defined distributed services that have "Intrinsic interoperability" are going to be beyond the grasp of a lot of IT shops both to develop and certainly to maintain/govern.  If you are going that route be sure to go over to Todd Biske's site - http://www.biske.com/blog/  and study up on governance (Going to need governance regardless, but for distributed "Intrinsic interoperability" you really going to need it).
 
Ultimately I believe the "Intrinsic interoperability" piece was put in to kind steer folks away from ESB's and the older EAI model of integration.  While I understand this from a pure academic view, I think in the actual wild it has issues.  It has the same issues that ESB's and EAI had because of lack of understanding of the methodology that went along with successful implementations. 
 
Overall the SOA-Manifesto is a good set of guiding principles.  Having a set of principles like this that everyone in the organization understands and embraces can go a long way to developing and maintaining a successful SOA style of architecture.

Friday, October 23, 2009

Windows 7

I upgraded to Windows 7 on my laptop yesterday.  Overall the upgrade went smooth but it was time consuming (About 3 hours, of course I have a lot of stuff on my laptop).  I think the average home user will probably not upgrade to Windows 7 but rather acquire it through a new PC purchase.  That's probably a good thing for the end user.
 
Here are few things I noticed during the upgrade.  The first thing that jumped out is that the upgrade process is still not geared towards the average end user.  The upgrade process is going to do a compatibility check before allowing the upgrade to proceed.  If it finds things that are not compatible it's going to ask you to uninstall them.  While not an issue for an IT person like myself, I suspect an average home user is now confused.  Especially when it asks you to uninstall your wireless network card.  I did like the fact that it tells you iTunes is probably not going to work correctly anymore.  Nice marketing play there Microsoft.  Of course it works fine after the upgrade.
 
On the positive side, Windows 7 takes up less memory and runs a bit faster.  On the downside they got rid of their little built in mail client which is a bummer.  I liked that client, really don't like web mail interfaces, sorry Google but I like to work disconnected sometimes.  Plus I don't like a million things on my screen other than the email when I working on email.  I know there are a million email clients out there to be had but I liked that one, it was simple and efficient.
 
It did detect all my hardware and configure it correctly which is a step up from the past.  The interface is clean and seems to be pretty logical.  Overall I'd give Windows 7 a thumbs up.  I'll post back any observations as I get more accustomed to it in the coming weeks.

Sunday, October 04, 2009

The Value of Enterprise Architecture

It's always kind of interesting when you have to explain and justify what seems like common sense. But like SOA, Enterprise Architecture seems to need constant justification and explanation. Why is that? Is it because we make what should be fairly straightforward concepts overly complicated? I tend to think so. Take a look at many of the EA frameworks available and your eyes will probably roll back in your head and you will lose consciousness.

I think keeping things simple tend to lead to more success. A major medical institution that will go unnamed spent millions of dollars on an automated medical scrubs dispensing system. It was going to save lots of money, folks wouldn't be allowed to take home and keep extra scrubs. After installing and testing the system, all things were go. Unfortunately there was one major flaw, a human being still had to load the machines each day with the clean scrubs.

As you might imagine, scrub machine loader is not the top end of the job spectrum. Scrubs are not one size fits all and each machine was responsible for dispensing the correct sizes. You can see this train wreck coming yes? After only a few months and tired of not ever being able to get the right size of scrubs, the employees figured out how to defeat the machines and keep out more scrubs than was suppose to be allocated. Millions spent and nothing gained.

When I was a young Army officer many years ago, I learned a very valuable lesson. Everyone needs to understand the mission. And by understanding it I mean be able to execute it. It seems like common sense doesn't it? If the leader becomes unable to perform his or her duties it falls to the next in line to carry out the mission.

In my simple world EA is also very simple. Define the destination, show everyone how to get there and then help them along the way to make sure they stay on course. Of course the downside to simplicity is that things don't tend to fail in such a grandiose way like when they are over engineered and over architected. Still it is important to realize that simplicity is not easy to achieve. Complex systems tend to form naturally.

Ultimately I think one of the keys to achieving simplicity is constant vigilance. How you get this working state of vigilance is the topic of another post. The hint is, there is not one answer for all.