Back when I was first getting into enterprise systems design at John Deere I remember taking a challenge we were having with me to TechEd to pose to the guys behind MTS (The precursor to COM+ which you probably know as Enterprise Services…). We were having challenges with performance moving around recordsets of data. I found the program manager for MTS and laid out my problem: It took 60 seconds to shuffle 45,000 rows from the MTS component on the server back to the client via DCOM. His response sticks with me to this day:
Son, any system that relies on moving forty five thousand rows between multiple computers to answer a user’s question has an architectural problem, not a performance problem.
At first I was taken aback: Our approach was dictated by the architectural rules handed to us by the enterprise architects, and seemed sound when drawn up in a nice Visio chart. Fortunately, the Microsoft Program Manager took the time to walk me through what we were doing at each stage, and just how bad the ratio of effective work (that ultimately used to provide value to our users) was to the total work being done. In fact, we had an architectural problem.
When we’re talking with prospective users of Gibraltar, we often run across folks that are doing virtually no logging at all in their applications. When they have a problem, they work to get it to replicate in Visual Studio and then fix it. Usually we can help them identify a few key places where they can add little logging and get a lot of coverage, like a common database access class, security system, etc. Occasionally however we can’t, and in these cases it almost always highlights a different problem: an ineffective architecture.
I didn’t like the others, they were all too flat.
I’m not advancing the theory that your architecture should be oriented around logging, but I do believe that a good architecture will have sufficient separation of concerns that you can find gateway points that interesting execution flows go through. If not, your architecture is likely:
- Too thin: You may be just providing a small veneer on top of someone else’s goliath that’s doing most of the work. In this case, hopefully that system either has hooks to extend it for monitoring or has a built in monitoring system you can activate.
- Too linear: Everything your app does has its own code from the outside interface of your application all the way down to the core.
The linear architecture is easy to get trapped into if you keep adding feature-by-feature following the typical samples published in MSDN or other simple scenarios. By their nature, these examples try to make a point in one area and minimize the work everywhere else. If you scaled them up you’d end up with data access all over your application in its controls, forms, everywhere. Each function would have little overlap with another function.
It’s seductive to keep adding each new feature with the same approach because you don’t have anything to build it on top of, and going back to create common reusable layers is a lot of work. Eventually if your application achieves any scale in terms of features it just won’t be maintainable.
Rapidly diagnose each error in any .NET application with our new Web Log Viewer and Exception root cause analysis, new in Loupe 4.5. New integration with Azure Service Bus and Azure Search enables full Loupe functionality without any Virtual Servers in Azure. Read more
The recently reported Cloudflare vulnerability where fragments of secure, encrypted user data could be exposed to a third party does not affect Gibraltar Software even though we use Cloudflare because we only route static content through the Cloudflare proxy for acceleration. Read more
Back in January of 2016 we decided to completely transition out of our data centers and into the cloud. On Sunday we finally shut down the last cluster of our hardware. Read more for how we did it and whether we would do it all over again if we had... Read more