Tomorrow morning, the Integrate 2014 conference revs up here in Redmond, WA with Microsoft presenting the vision for the next versions of BizTalk Server and Microsoft Azure BizTalk Services.
I’m getting pretty excited as I’m looking over the first day agenda to see what to expect from the talks and sessions like Business Process Management and Enterprise Application Integration placed in between talks about BizTalk Services and On-Premise BizTalk Server. I’m even more excited seeing the second day agenda and seeing talks like Business Rules Improvements, and especially Building Connectors and Activities.
What’s Keeping Me Busy
In anticipation of this event, I have been building out a hybrid cloud integration sample application that leverages some of the ideas laid out in my previous post regarding cloud integration patterns while also providing the content for the long overdue next piece in my new features series for BizTalk Server 2013 R2.
In the sample, I’m running requests through MABS for content-based routing based on an external SQL lookup, with the requests targeting either an on-premise SQL Server database or a BizTalk Server instance for further processing that can’t otherwise be done currently with BizTalk Services (namely Business Rules Evaluation and Process Orchestration).
I’m hoping that after the conference this week, I will be able to tear the sample apart and build out most (if not all) of the elements in the cloud.
Best Bridge Between Azure and On-Prem?
Along the way, I’ve been further playing with using Service Bus Topics/Subscriptions as the Message Box of the cloud. At the same time, it represents a nice bridge between BizTalk Services and On-Premise BizTalk Server.
Consider the itinerary below;
This was actually the first draft of prototyping out the application. What this represents is an application that is receiving requests from client applications that have been published to a service bus topic. As they are received, they are routed based on content and sent either to a database to be recorded (as an example of an on-premise storage target), or to BizTalk Server 2013 R2 for further processing through a Service Bus Relay.
Given the scenario, maybe a relay is appropriate – lower latency requirement, no requirement for durability (which we have already sacrificed by running it through an initial bridge).
However, maybe we want to take a more holistic approach, and assume that the cloud is giving us a free public endpoint and some quite powerful content-based routing, translation, and even publish-subscribe capability when we bring Azure Service Bus to the mix. Let’s further assume that we view these capabilities as simply items in our toolbox alongside everything that BizTalk Server 2013 R2 is already providing us.
Let’s make it more concrete. What if the on-premise processing ultimately ends up sending the message back to the TradesDb? Is there a potential waste in building out that portion of the process in both locations?
Service Bus is the New Hybrid Integration MessageBox
Let’s try this instead:
Here, instead of using a relay, we’re using a Service Bus Topic to represent the call out to BizTalk Server 2013 R2.
Why would we do this? While it introduces slightly more latency (in theory – though I haven’t properly tested that theory), it becomes pure loosely coupled pub-sub. I’m going to go out on a limb here (and risk being called an architecture astronaut), and say that not only is that not a bad thing, but it might even be a good idea. By feeding a topic rather than directly submitting to BizTalk Server via Relay allows us to easily swap out the processing of the rules with any mechanism we want, at any time – even if it means that we will have to transform the message and further submit by a completely different transport. Maybe one day we will be able to replace this with a call to a Rules Engine capability within MABS (crossing my fingers here) if we see such a capability come.
Further, we have broken out the logging of trades to the database into it’s own separate miniature process alongside the rest. This one might be fed by messages generated by BizTalk Server 2013 R2 on-premise or the earlier processing in MABS – providing only a single implementation of the interface to manage and change.
Is It The Right Way™?
I don’t know. It feels kind of right. I can see the benefits there, but again, we are potentially making a sacrifice by introducing latency to pay for a loosely coupled architecture. Call me out on it in the comments if that makes you unhappy. Ultimately, this would have to simply become a consideration that is weighed in choosing or not choosing to do things this way.
What Could Make it Even Better?
Imagine a time when we could build-out a hybrid integration using a design surface that makes it seamless. One where I could quickly discover the rest of the story. Right now there’s quite a bit happening on-premise into which we have no visibility via the itinerary – and very limited visibility within the orchestration since most logic is in Business Rules and my maps happen just before/after port processing.
Tomorrow I will be writing a follow-up blog with a re-cap of the first day of the Integrate 2014 conference. Additionally, I will be playing with this application in my mind and seeing where the things announced this week change the possibilities.
If you’re going to be there tomorrow, be sure to stop by the QuickLearn Training table to sign-up for a chance to win some fun prizes. You can also just stop by to talk about our classes, or about any of the ideas I’ve thrown out here – I welcome both the positive and negative/nitpicky feedback.
Also, make sure you’re following @QuickLearnTrain on Twitter. We’ll be announcing an event that you won’t want to miss sometime in the next few days.
See you at the conference!
– Nick Hauenstein