Fundamentally, as companies increasingly adopt more and more SaaS/cloud-based applications, as more and more data are cloud-based, and as social networking data becomes increasingly critical to sales, marketing and customer satisfaction applications, the "old style" integration stacks that were originally created to run on on-premises PCs or servers will be a non-optimal solution.
This is the normal evolution of things. The mainframe-centric nature of computing gave way to client/server which in turn gave way to Internet and then SaaS/cloud.
Just like mainframe-based applications could never really be retrofitted to be good client/server applications, or how non-relational databases never really worked well with SQL front-ends slapped on, older architectures (server or LAN-based application integration suites) can't simply be retrofitted or wrappered to become sky-based.
As the nature of applications and data change, so must the nature of the integration suites that bring them all together. The EDI hub from 1990 isn't going to provide the right architecture for this. The B2B or EAI server from 2001 can't do this. The nature of the problem has changed.
I'm not saying that every bit of "old style" integration technology needs to be thrown out. A lot of the technologies are similar. But the changing nature of data, applications and integration is significant enough to drive important changes in the integration suites that bring all this data together. Especially for those organizations who plan on (or already have) a lot of cloud/SaaS-based data.
Perhaps most important, though, is the substantial leap forward in productivity and ease of use that the new cloud-based integration suites bring to the table - so irrespective of where your data are, you may want to consider a cloud-based solution for that reason alone.
The Locus of Integration
Apart from the things that I've already mentioned, there is one more important thing that older architectures aren't adept at handling: the changing locus of integration.
Many years ago, the locus, or "center of effort" for integration was the mainframe. Simply because every application and every bit of data were mainframe-based.
In the 1990s and early 2000s when many integration stacks were first built, the locus of integration moved to "the data center", as applications were mostly distributed on servers based in the data center, and connected via the LAN/WAN".
As an increasing number of applications and data silos are sky (i.e., cloud and/or SaaS) based, the locus of integration is shifting closer and closer to the sky. For many organizations, this locus is ALREADY closer to the cloud than it is to the data center.
At the point where the locus of integration becomes closer to the sky than the ground, it makes sense to consider moving the ACT of integration itself (i.e. the Integration hub) into the Cloud.
So, one key aspect of Integration for 2012 and beyond is the ability to support the changing locus of integration by supporting data center-centric, cloud-centric data and applications.
A Word to the Wise
Be careful - there are fair number of "pretenders" out there "technology comb-overs" that pretend to be cloud or SaaS solutions, but are not the "real thing." Just because something runs "on a cloud" doesn't make it a cloud or SaaS solution.
One of the primary benefits of the new "SaaS / cloud" world is so much about enhanced productivity. That doesn't just happen magically by stuffing a bunch of decades-old EDI or ETL code up on the cloud and running it there. You might as well take a COBOL program from 1975 and somehow make it work on Amazon's Cloud - character based "green screen" interface and all.
Cloud and SaaS Integration products need to be about a modern event-based, services-based integration architecture - based on (not just supporting) approaches such as REST or SOAP. Multi-tenant, pay-as-you-go, the whole thing. And they need to have modern tools for development, configuration and deployment - all browser based.
The new generation of cloud-based Integration solutions such as Boomi (Dell), InformaticaCloud, SnapLogic and newcomer MuleSoft iON (among others) are designed from the ground up to be cloud-based integration platforms. These REAL cloud-based products have implementation/deployment times measured in a few weeks.
Many of the "re-tread" pretenders are just the same "dinosaur" on-premises EDI, EAI or data integration solutions that have been around since before people realized that Y2K was a problem. These are products that have a reputation for taking 3-6 months (or longer) to implement - and for years have given Integration a bad name for reportedly being complex, risky, brittle, and difficult/expensive to implement.
Cloud-Based Integration for Cloud-based Applications
Anyhow, it should come as no surprise to you that that 95% of companies (based on a recent survey) plan to implement cloud-based integration. Over 40% of respondents indicated that cloud-based integration will be part of their future within two years.
I was personally a bit baffled by 5% stating that cloud-based integration would "never" be implemented at their organizations. I suspect (although do not know) that those organizations have certain characteristics that do not lend themselves to the movement of sensitive data outside carefully protected boundaries. Or, as David Linthicum pointed out recently, organizations such as the Federal Government - who, although they acknowledge the cost savings of cloud over time, they lack the seed capital to make the move "right now".
Conclusion and Recommendations
It's clear that the very nature of applications and data - where they reside, what forms they are in, how and by whom they are used, how often they change, etc. is in a state of rapid transformation.
The very nature of application and data integration needs to change at an architectural level in order to accommodate the corresponding architectural-level shift in applications and data. To date, many of the traditional application integration technologies have failed (in varying degrees) to rise to that challenge. Their older LAN or Server-based architectures will leave them increasingly at a competitive disadvantage.
It would behoove those choosing cloud-based integration solutions to be careful in choosing cloud-based integration platforms to make sure that they're getting the real deal. Otherwise, they may find themselves giving answers like "cost too much," "took too long," "wasn't flexible enough" in some future market research survey.
Note: The data referred to in this article come from 2011 Application Connection Priorities a research report by GatePoint Research.
Subscribe to SOA World Magazine Email News Alerts
Subscribe via RSS
