Architecture at different levels of abstraction

This week I am attending the O'Reilly Architecture Conference in NYC with my team and I had the pleasure of attending a two day training on Building Evolutionary Architectures presented by Neal Ford and Rebecca Parsons.

During this session, on of the things we talked about extensively was coupling. We looked at dependency maps for different architectural styles including the big ball of mud. We saw something like the following and almost everyone agreed, "That is a bad architecture."

Big Ball of Mud

This reminded me of something I has seen elsewhere before, and after a quick search, I found it:


Amazon and Netflix Death Star Architectures

The ball of mud is considered obviously wrong by most architects, but the death star seems to be held up as an ideal by just as many and likely most of the same architects. Why is this?

  • Are afferent coupling (inbound dependencies) and efferent coupling (outreaching dependencies) only important at a certain scale?
  • Is it a matter of getting things done? We feel more productive when we are doing something, and reports suggest that individual teams are regularly delivering software at Netflix. (Are they delivering value?) 
  • Is it more obvious that teams are struggling to deliver value in a smaller scope? At a larger scope, we can see a lot of activity so the investment required to get progress seems more acceptable?
  • Maybe we're impressed by the tooling that Netflix has created to manage this death star?
  • In the end is this a case of bike shedding?
I am not trying to indict a major media company for their choices. I happily pay them my monthly subscription for the entertainment they bring into my house. I want to know what makes these design choices wrong at one scale but somehow right at another scale. Is this the architecture you would choose for your system?

Comments

Post a Comment

Popular posts from this blog

TFS to SVN Conversion with History

Recently, the team I was on wanted to move our source code from Team Foundation Server to SVN for reasons that had almost nothing to do with source control directly. This would be a pretty simple operation, but we didn't want to lose access to the history when the operations team turned off the machine running TFS. We couldn't find any direct paths for moving history from TFS to SVN, so instead we chose to use git as an intermediary. Here's how we did it: Step 1 - Install the tools Install msysgit: http://code.google.com/p/msysgit/ Install git-tfs: https://github.com/git-tfs/git-tfs Step 2 - Pull TFS history to local git repository git tfs clone <tfs-project-url> "$/<tfs-project-name>/<tfs-project-folder>" <localfolder> example: git tfs clone http://tfs-001:8080/tfs/projectroot "$/Reference Architecture/ReferenceArchitecture" RefArch Step 3 - Initialize git-svn * User Lower Case - Caps will make it fail! * Change direc
Read more

Simpler Tests: What kind of test are you writing?

One common problem that I see in test suites is confusion about what each test should cover. This confusion often leads to tests that are either too broad or too focused to accomplish the goal of the author. When writing a test, it is important to think about what kind of test it will be and the constraints that make that type of test effective. There are a few broad categories of tests that I keep in mind to help focus my testing: Unit Tests Also called micro tests, unit tests form the foundation of any good testing strategy. These are tests that test a single class or method in isolation from it's dependencies. Mocks and stubs help us isolate the code under test and make our tests more focused and expressive. A test that touches any external resource such as the database, the file system, a web service, etc is never a unit test. The primary purpose of unit tests is to validate that the subject under test is functionally correct for the expected inputs and outputs. Additiona
Read more