Subscribe to Methods & Tools
if you are not afraid to read more than one page to be a smarter software developer, software tester or project manager!
Software Development Blogs: Programming, Software Testing, Agile Project Management
Subscribe to Methods & Tools
if you are not afraid to read more than one page to be a smarter software developer, software tester or project manager!
Here's a little snippet that my class really picked up on yesterday. During the training course, we get people into groups and ask them to design a solution based upon some simple requirements. The deliverable is "one or more diagrams to describe your solution". Aside from answering a few questions about the business domain and the environment, that's pretty much all the guidance that groups get.
As you can probably imagine, the resulting diagrams are all very different. Some are very high-level, others very low-level. Some show static structure, others show runtime and behavioural views. Some show technology, others don't. Without a consistent approach, these differing diagrams make it hard for people to understand the solutions being presented to them. But furthermore, the differing diagrams make it really hard to compare solutions too.
As I've said before, I don't actually teach people to draw pictures. What I do instead is to teach people how to think about, and therefore describe, their software using a simple set of abstractions. This is my C4 model. With these abstractions in place, groups then redraw their diagrams. Despite the notations still differing between the groups, the solutions are much easier to understand. The solutions are much easier to compare too, because of the consistency in the way they are being described. A common set of abstractions is much more important than a common notation.
I've been writing blog posts covering a number of topics over the past few months; from the conflict between software architecture and code and architecturally-evident coding styles through to representing a software architecture model as code and how microservice architectures can easily turn into distributed big balls of mud. The common theme running throughout all of them is structure, and this in turn has a relationship with testability.
The TL;DR version of this post is: think about modularity, think about how you structure your code, think about the options you have for testing your code and stop making everything public.1. The conflict between software architecture and code
I've recently been talking a lot about the disconnect between software architecture and code. George Fairbanks calls this the "model-code gap". It basically says that the abstractions we consider at the architecture level (components, services, modules, layers, etc) are often not explicitly reflected in the code. A major cause is that we don't have those concepts in OO programming languages such as Java, C#, etc. You can't do public component X in Java, for example.2. The "unit testing is wasteful" thing
Hopefully, we've all see the "unit testing is wasteful" thing, and all of the follow-up discussion. The unfortunate thing about much of the discussion is that "unit testing" has been used interchangeably with "TDD". In my mind, the debate is about unit testing rather than TDD as a practice. I'm not a TDDer, but I do write automated tests. I mostly write tests afterwards. But sometimes I write them beforehand, particularly if I want to test-drive my implementation of something before integrating it. If TDD works for you, that's great. If not, don't worry about it. Just make sure that you *do* write some tests. :-)
There are, of course, a number of sides to the debate, but in TDD is dead. Long live testing. (ignore the title), DHH makes some good points about the numbers and types of tests that a system should have. To quote (strikethrough mine):I think that's the direction we're heading. Less emphasis on unit tests, because we're no longer doing test-first as a design practice, and more emphasis on, yes, slow, system tests. (Which btw do not need to be so slow any more, thanks to advances in parallelization and cloud runner infrastructure).
The type of software system you're building will also have an impact on the number and types of tests. I once worked on a system where we had a huge number of integration tests, but very few unit tests, primarily because the system actually did very little aside from get data from a Microsoft Dynamics CRM system (via web services) and display it on some web pages. I've also worked on systems that were completely the opposite, with lots of complex business logic.
There's another implicit assumption in all of this ... what's the "unit" in "unit testing"? For many it's an isolated class, but for others the word "unit" can be used to represent anything from a single class through to an entire sub-system.3. The microservices hype
Microservices is the new, shiny kid in town. There *are* many genuine benefits from adopting this style of architecture, but I do worry that we're simply going to end up building the next wave of distributed big balls of mud if we're not careful. Technologies like Spring Boot make creating and deploying microservices relatively straightforward, but the design thinking behind partitioning a software system into services is still as hard as it's ever been. This is why I've been using this slide in my recent talks.
Uncle Bob Martin posted Microservices and Jars last month, which touches upon the topic of building monolithic applications that do have a clean internal structure, by using the concept of separately deployable units (e.g. JARs, DLLs, etc). Although he doesn't talk about the mechanisms needed to make this happen (e.g. plugin architectures, Java classloaders, etc), it's all achievable. I rarely see teams doing this though.
Structuring our code for modularity at the macro level, even in monolithic systems, provides a number of benefits, but it's a simple way to reduce the model-code gap. In other words, we structure our code to reflect the structural building blocks (e.g. components, services, modules) that we define at the architecture level. If there are "components" on the architecture diagrams, I want to see "components" in the code. This alignment of architecture and code has positive implications for explaining, understanding, maintaining, adapting and working with the system.
It's also about avoiding big balls of mud, and I want to do this by enforcing some useful boundaries in order to slice up my thousands of lines of code/classes into manageable chunks. Uncle Bob suggests that you can use JARs to do this. There are other modularity mechanisms available in Java too; including SPI, CDI and OSGi. But you don't even need a plugin architecture to build a structured monolith. Simply using the scoping modifiers built in to Java is sufficient to represent the concept of a lightweight in-process component/module.Stop making everything public
We need to resist the temptation to make everything public though, because this is often why codebases turn into a sprawling mass of interconnected objects. I do wonder whether the keystrokes used to write public class are ingrained into our muscle memory as developers. As I said during my closing session at DevDay in Krakow last week, we should make a donation to charity every time we type public class without thinking about whether that class really needs to be public.
A simple way to create a lightweight component/module in Java is to create a public interface and keep all of the implementation (one or more classes) package protected, ensuring there is only one "component" per package. Here's an example of a such a component, which also happens to be a Spring Bean. This isn't a silver bullet and there are trade-offs that I have consciously made (e.g. shared domain classes and utility code), but it does at least illustrate that all code doesn't need to be public. Proponents of DDD and ports & adapters may disagree with the naming I've used but, that aside, I do like the stronger sense of modularity that such an approach provides.Testability
And now you have some options for writing automated tests. In this particular example, I've chosen to write automated tests that treat the component as a single thing; going through the component API to the database and back again. You can still do class-level testing too (inside the package), but only if it makes sense and provides value. You can also do TDD; both at the component API and the component implementation level. Treating your components/modules as black boxes results in a slightly different testing pyramid, in that it changes the balance of class and component tests.
A microservice architecture will likely push you down this route too, with a balanced mix of low-level class and higher-level service tests. Of course there is no "typical" shape for the testing pyramid; the type of system you're building will determine what it looks like. There are many options for building testable software, but neither unit testing or TDD are dead.
In summary, I'm looking for ways in which it we can structure our code for modularity at the macro-level, to avoid the big ball of mud and to shrink the model-code gap. I also want to be able to automatically draw some useful architecture diagrams based upon the code. We shouldn't blindly be making everything public and writing automated tests at the class level. After all, there are a number of different approaches that we can take for all of this, and the modularity you choose has an implication on the number and types of tests that you write. As I said at the start; think about modularity, think about how you structure your code, think about the options you have for testing your code and stop making everything public. Designing software requires conscious effort. Let's not stop thinking.
I truly believe that a lightweight, pragmatic approach to software architecture is pivotal to successfully delivering software, and that it can complement agile approaches rather than compete against them. After all, a good architecture enables agility and this doesn't happen by magic. But the people in our industry often tend to have a very different view. Historically, software architecture has been a discipline steeped in academia and, I think, subsequently feels inaccessible to many software developers. It's also not a particularly trendy topic when compared to [microservices|Node.js|Docker|insert other new thing here].
I've been distilling the essence of software architecture over the past few years, helping software teams to understand and introduce it into the way that they work. And, for me, the absolute essence of software architecture is about building firm foundations; both in terms of the team that is building the software and for the software itself. It's about technical leadership, creating a shared vision and stacking the chances of success in your favour.
I'm delighted to have been invited back to ASAS 2014 and my opening keynote is about what a software team can do in order to create those firm foundations. I'm also going to talk about some of the concrete examples of what I've done in the past, illustrating how I apply a minimal set of software architecture practices in the real world to take an idea through to working software. I'll also share some examples of where this hasn't exactly gone to plan too! I look forward to seeing you in a few weeks.
My Software architecture vs code talk has evolved considerably over the past year and I've presented a number of iterations at events in Europe and the US. If you're interested, thanks to the nice people behind the GOTO conferences, you can now watch the video of the version that I presented at GOTO Chicago 2014.
It needs a little updating (isn't that always the case!), but I've moved the example software guidebook (previously an appendix in my Software Architecture for Developers book) into a separate free and open source book on Leanpub.
techtribes.je is a side-project of mine to create a content aggregator for the tech, IT and digital sector in Jersey, Channel Islands. The code behind the techtribes.je website is open source and available on GitHub. The source for the software guidebook is also open source and available on GitHub.
The techtribes.je software guidebook is based upon the concept of a software guidebook as described in my Software Architecture for Developers book; the software guidebook is a lightweight, pragmatic way to document the "big picture" of a software system. In essence, it's my simplified version of many "software architecture document" templates you'll find out there on the web.
techtribes.je - Software Guidebook is available to download for free from Leanpub. I hope you find it useful.
There's been some interesting discussion over the past fews days about Leanpub, both on Twitter and blogs. Jurgen Appelo posted Why I Don't Use Leanpub and Peter Armstrong responded. I think the biggest selling points of Leanpub as a publishing platform from an author's perspective may have been lost in the discussion. So, here's why my take on why I use Leanpub for Software Architecture for Developers.Some history
I pitched my book idea to a number of traditional publishing companies in 2008 and none of them were very interested. "Nice idea, but it won't sell" was the basic summary. A few years later I decided to self-publish my book instead and I was about to head down the route of creating PDF and EPUB versions using a combination of Pages and iBooks Author on the Mac. Why? Because I love books like Garr Reynolds' Presentation Zen and I wanted to do something similar. At first I considered simply giving the book away for free on my website but, after Googling around for self-publishing options, I stumbled across Leanpub. Despite the Leanpub bookstore being fairly sparse at the start of 2012, the platform piqued my interest and the rest is history.The headline: book creation, publishing, sales and distribution as a service
I use Leanpub because it allows me to focus on writing content. Period. The platform takes care of creating and selling e-books in a number of different formats. I can write some Markdown, sync the files via Dropbox and publish a new version of my book within minutes.Typesetting and layout
I frequently get asked for advice about whether Leanpub is a good platform for somebody to write a book. The number one question to ask is whether you have specific typesetting/layout needs. If you want to produce a "Presentation Zen" style book or if having control of your layout is important to you, then Leanpub isn't for you. If, however, you want to write a traditional book that mostly consists of words, then Leanpub is definitely worth taking a look at.
Leanpub uses a slightly customised version of Markdown, which is a super-simple language for writing content. Here's an example of a Markdown file from my book, and you can see the result in the online sample of my book. Leanpub does allow you to tweak things like PDF page size, font size, page breaking, section numbering, etc but you're not going to get pixel perfect typesetting. I think that Leanpub actually does a pretty fantastic job of creating good looking PDF, EPUB and MOBI format ebooks based upon the very minimal Markdown. This is especially true when you consider the huge range of ebook reader software across PCs, Macs, Android devices, Apple devices, Kindles, etc. Plus the readers themselves can mess with the fonts/font sizes too.
It's like building my own server at Rackspace versus using a "Platform as a Service" such as Cloud Foundry. You need to make a decision about the trade-off between control and simplicity/convenience. Since authoring isn't my full-time job and I have lots of other stuff to be getting on with, I'm more than happy to supply the content and let Leanpub take care of everything else for me.Toolchain
My toolchain as a Leanpub author is incredibly simple: Dropbox and Mou. From a structural perspective, I have one Markdown file per essay and that's basically it. Leanpub does now provide support for using GitHub to store your content and I can see the potential for a simple Leanpub-aware authoring tool, but it's not rocket science. And to prove the point, a number of non-technical people here in Jersey have books on Leanpub too (e.g. Thrive with The Hive and a number of books by Richard Rolfe).Iterative and incremental delivery
Before starting, I'd already decided that I'd like to write the book as a collection of short essays and this was cemented by the fact that Leanpub allows me to publish an in-progress ebook. I took an iterative and incremental approach to publishing the book. Rather than starting with essay number one and progressing in order, I tried to initially create a minimum viable book that covered the basics. I then fleshed out the content with additional essays once this skeleton was in place, revisiting and iterating upon earlier essays as necessary. I signed up for Leanpub in January 2012 and clicked the "Publish" button four weeks later. That first version of my book was only about ten pages in length but I started selling copies immediately.Variable pricing and coupons
Another thing that I love about Leanpub is that it gives you full control over how you price your book. The whole pricing thing is a balancing act between readership and royalties, but I like that I'm in control of this. My book started out at $4.99 and, as content was added, that price increased. The book now currently has a minimum price of $20 and a recommended price of $30. I can even create coupons for reduced price or free copies too. There's some human psychology that I don't understand here, but not everybody pays the minimum price. Far from it, and I've had a good number of people pay more than the recommend price too. Leanpub provides all of the raw data, so you can analyse it as needed.An incubator for books
As I've already mentioned, I pitched my book idea to a bunch of regular publishing companies and they weren't interested. Fast-forward a few years and my book is the currently the "bestselling" book on Leanpub this week, fifth by lifetime earnings and twelfth in terms of number of copies sold. I've used quotes around "bestselling" because Jurgen did. ;-)
In his blog post, Peter Armstrong emphasises that Leanpub is a platform for publishing in-progress ebooks, especially because you can publish using an iterative and incremental approach. For this reason, I think that Leanpub is a fantastic way for authors to prove an idea and get some concrete feedback in terms of sales. Put simply, Leanpub is a fantastic incubator for books. I know of a number of books that were started on Leanpub have been taken on by traditional publishing companies. I've had a number of offers too, including some for commercial translations. Sure, there are other ways to publish in-progress ebooks, but Leanpub makes this super-easy and the barrier to entry is incredibly low.The future for my book?
What does the future hold for my book then? I'm not sure that electronic products are ever really "finished" and, although I consider my book to be "version 1", I do have some additional content that is being lined up. And when I do this, thanks to the Leanpub platform, all of my existing readers will get the updates for free.
I've so far turned down the offers that I've had from publishing companies, primarily because they can't compete in terms of royalties and I'm unconvinced that they will be able to significantly boost readership numbers. Leanpub is happy for authors to sell their books through other channels (e.g. Amazon) but, again, I'm unconvinced that simply putting the book onto Amazon will yield an increased readership. I do know of books on the Kindle store that haven't sold a single copy, so I take "Amazon is bigger and therefore better" arguments with a pinch of salt.
What I do know is that I'm extremely happy with the return on my investment. I'm not going to tell you how much I've earned, but a naive calculation of $17.50 (my royalty on a $20 sale) x 4,600 (the total number of readers) is a little high but gets you into the right ballpark. In summary, Leanpub allows me focus on content, takes care of pretty much everything and gives me an amazing author royalty as a result. This is why I use Leanpub.
After a lovely summer (mostly) spent in Jersey, September is right around the corner and is shaping up to be a busy month. Here's a list of the events where you'll be able to find me.
It's going to be a fun month and besides, I have to keep up my British Airways frequent flyer status somehow, right? ;-)
A few people have recently asked me for a poster/cheat sheet/quick reference of the C4 model that I use for communicating and diagramming software systems. You may have seen an old copy floating around the blog, but I've made a few updates and you can grab the new version from http://static.codingthearchitecture.com/c4.pdf (PDF, A3 size).
I was reading Dan North Visits Osper and was pleasantly surprised to see Dan mention CRC modelling. CRC is a great technique for the process of designing software, particularly when used in a group/workshop environment. It's not a technique that many people seem to know about nowadays though.
A Google search will yield lots of good explanations on the web, but basically CRC is about helping to identify the classes needed to implement a particular feature, use case, user story, etc. You basically walk through the feature from the start and whenever you identify a candidate class, you write the name of it on a 6x4 index card, additionally annotating the card with the responsibilities of the class. Every card represents a separate class. As you progress through the feature, you identify more classes and create additional cards, annotating the cards with responsibilities and also keeping a note of which classes are collaborating with one another. Of course, you can also refactor your design by splitting classes out or combining them as you progress. When you're done, you can dry-run your feature by walking through the classes (e.g. A calls B to do X, which in turn requests Y from C, etc).
Much of what you'll read about CRC on the web discusses how the technique is useful for teaching OO design at the class level, but I like using it at the component level when faced with architecting a software system given a blank sheet of paper. The same principles apply, but you're identifying components (or services, microservices, etc) rather than classes. When you've done this for a number of significant use cases, you end up with a decent set of CRC cards representing the core components of your software system.
From this, you can start to draw some architecture diagrams using something like my C4 model. Since the cards represent components, you can simply lay out the cards on paper, draw lines between them and you have a component diagram. Each of those components needs to be running in an execution environment (e.g. a web application, database, mobile app, etc). If you draw boxes around groups of components to represent these execution environments, you have a containers diagram. Step up one level further and you can create a simple system context diagram to show how your system interacts with the outside world. My Simple Sketches for Diagramming your Software Architecture article provides more information about the C4 model and the resulting diagrams, but hopefully you get the idea.
CRC then ... yes, it's a great technique for collaborative design, particularly when applied at the component level. And it's a nice starting point for creating software architecture diagrams too.
In Diagramming Spring MVC webapps, I presented an approach that allows you to create a fairly comprehensive model of a software system in code. It starts with you creating a simple base model that includes software systems, people and containers. With this in place, all of the components can then be automatically populated into the model via a scan of the compiled Java code. This is all based upon Software architecture as code.
Once you have a model to work with, it's relatively straightforward to visualise it via a number of views. In the Spring PetClinic example, three separate views (one each of a system context, containers and components view) are sufficient to show everything. With larger software systems, however, this isn't the case.
As an example, here's what a single component diagram for the web application of my techtribes.je system looks like.
Yup, it's a mess. The components around the left, top and right edges are Spring MVC controllers, while those in the centre are the core components. There are clearly three hotspots here - the LoggingComponent, ActivityComponent and ContentSourceComponent. The reason for the first should be obvious, in that almost all components use the LoggingComponent. The latter two are used by all controllers, simply because some common information is displayed on the header of all pages on the website. I don't mind excluding the LoggingComponent from the view, but I'd quite like to keep the other two. That aside, even excluding the ActivityComponent and ContentSourceComponent doesn't actually solve the problem here. The resulting diagram is still a mess because it's showing far too much information. Instead, another approach is needed.
With this in mind, what I've done instead is use a programmatic approach to create a number of views for the techtribes.je web application, one per Spring MVC controller. The code looks like this.
The result is a larger number of simple diagrams, but I think that the trade-off is worth it. It's a much better way to navigate a large model.
And here's an example component diagram that focusses on a single Spring MVC controller.
The JSON representing the techtribes.je model can be found on GitHub and you can copy-paste it into my (still in-progress) diagramming tool if you'd like to explore the model yourself. I'm still experimenting with much of this but I really like the opportunities provided by having the software architecture model in code. This really is "software architecture for developers". :-)
Simon recently talked about the gap between Software Architecture and Code and how to close this with architecturally-evident coding. He's also creating tools to allow Software Architecture to be expressed as code.
If you're working on a greenfield project then including annotations to help with navigation is a great solution but what if you've inherited a large system with a model-code gap? Or if you only realise, sometime into a project, that you lack a model to help you understand its growing complexity? Well, Simon also had some thoughts on scanning Spring annotations to provide this data. This works quite well and it got me thinking about other artifacts in code that can be extracted for these diagrams.
(In more formal terms - for those of you that like to quote ISO42010 - we are trying to extract architectural elements from code that can be displayed within architectural views. Of course the elements may be from a variety of differing abstractions, levels and granularity and therefore need to be placed within differing views.)
So what can we extract from a current/legacy system to give us a view into it? Some suggestions include:Annotations As already suggested, dependency injections systems such as Spring provide some annotations that can be extracted to give a basic, high level model. Annotations are also present in Java EE applications and other enterprise frameworks. XML DI Configuration files Many (legacy) Spring projects use xml configuration files to define beans. Having scanned a few examples this seems to create a relatively low level model which would need some manual tweaking after generation. With sensible naming convention for beans you can produce models for a desired abstraction. The bean properties indicate the connections between these elements. Module Bundling Systems Modular systems such as OSGi define bundles of components and services including lifecycle and service registry. The deployment information should provide a high level overview. Packages If you have used 'package-by-component' then your packages will relate one-to-one with your components. The links between components should be identifiable by the links between the classes within them (the has-a relationships). If you have package-by-layer then this is much harder or impossible to use. Experience tells me that most real-world systems are actually a combination of the two so you should have some useful information. Class Names It's very common for class names to contain a strong indication of their role e.g. XyzService, XyzConnector, XyzDao, XyzFacade etc. Scanning for known patterns should identify the element names and roles. Interfaces and class hierarchies If you implement interfaces (or extend base classes) then the interfaces used may show the abstraction level and type e.g. implementing Service, DAO, Connector, Repository etc Delegation or Library Dependency Shared delegates used by a set of classes/functions may indicate their purpose. e.g. components delegating to a database utility might indicate a DAO component or using a CORBA utility might indicate a service. This is likely to be time consuming as you need to identify and scan for each delegate you are interested in. Comments/Javadoc/JSDoc/NDoc/Doclets Comments and javadoc style API comments can provide a large amount of meta-information about a class or package. In fact, many UML modeling tools enrich code using custom comments and tags. This has the advantage of not affecting the compiled code or introducing library dependencies but may not be consistently used. Tests Test can provide a lot of meta-data about your system. Unit tests tend to be concentrated around important classes and often construct entire components to test. Simply extracting the names of classes that are directly tested will produce a useful list of components. The higher level systems tests will reveal the important services. Build Systems Build systems such as ant, maven, NuBuild etc all have hooks into the code base for building and deployment. A simple extraction of the build targets will give you the deployment modules (which is a very helpful view for operation teams). This may give you the required information for a Containers view.
What do you think? Of course all of the above is very dependent on your codebase but if none of the above works then you have to question the quality and structure of the code! The data extracted may need filtering and manual correction as it won't give you exactly what you want. You might consider creating structurizr annotations using an initial scan and then maintaining them. One of my tasks for the next few weeks is to try this out on some legacy codebases I maintain.
What other ways of identifying architectural elements can you think of?
Regular readers will know that I'm a big fan of pictures, especially as a mechanism for communicating the structure of software systems. To this end, I sometimes introduce myself as somebody who teaches people how to draw pictures. This is often said with a smile, because on the face of it this is exactly what I appear to be doing. But there's less truth to this than there initially appears.
Something that we do on the full 2-day training course and the 1-day sketching workshop is to get people drawing some pictures to communicate the structure of a software system. This is either a solution to the financial risk system or their own software. Over the years, I've done this for thousands of people and pretty much everybody struggles to draw something that effectively communicates the software. The same challenges crop up again and again.
The diagrams that are produced during the initial 90-minute timebox are usually fairly confusing and exhibit problems ranging from a lack of clarity around the abstractions and notation used through to diagrams that show too little or too much detail. This is an aside, but the majority of the diagrams are informal "boxes and lines" rather than UML.
After some reviews and feedback, I introduce people to my C4 model. In a nutshell, it's all about thinking of a software system as being made up of containers (web applications, databases, mobile devices, etc), each of which contains components, which in turn are made up of classes. There are some variations of this depending on the technology you're using, but that's basically it. With this structure in mind, you can then draw a diagram at each level in turn, which leads to a system context diagram, a container diagram, one or more component diagrams and (optionally) some class diagrams. If you want a slightly more detailed introduction to C4, take a look at Simple sketches for diagramming your software architecture.
Back to the workshops, and iteration two is another 90-minute timebox in which teams can redraw their diagrams. I don't actually mandate that people should follow my C4 approach, but most do, and what happens next is really interesting. The conversations change. In stark contrast to the first timebox, the conversations are much more technical. Gone are the discussions about what to include on each diagram and instead the focus is on the technical aspects of the software. Instead of "what should we show on this diagram?", the questions I hear are more along the lines of "what are the responsibilities of this thing?" and "how does this thing talk to that thing?". In essence, the discussion returns to be about software design, which is exactly where it should be.
There are a number of ways in which you can communicate the architecture of a software system, so I don't want to pitch my C4 approach as "the one true way". What I do want to say is that providing even a little guidance in this area can go a long way. I ask people for their thoughts after the second timebox and the consensus is that the diagrams are easier to draw and comprehend because of the "framework" (their words, not mine) that's been provided. All I'm doing is providing some constraints that people might want to work within, and this frees them up to focus on what's important again.
Back to that comment about introducing myself as somebody who teaches people how to draw pictures. It turns out that this isn't actually what I do. I don't really care so much *how* people draw pictures. What I do care about is that they use a well-understood, consistent and meaningful set of *abstractions* to think about and therefore describe their software. It turns out that modelling our software isn't dead after all. ;-)
If you want evidence that the software development industry is susceptible to fashion, just go and take a look at all of the hype around microservices. It's everywhere! For some people microservices is "the next big thing", whereas for others it's simply a lightweight evolution of the big SOAP service-oriented architectures that we saw 10 years ago "done right". I do like a lot of what the current microservice architectures are doing, but it's by no means a silver bullet. Okay, I know that sounds obvious, but I think many people are jumping on them for the wrong reason.
I often show this slide in my conference talks, and I've blogged about this before, but basically there are different ways to build software systems. On the one side we have traditional monolithic systems, where everything is bundled up inside a single deployable unit. This is probably where most of the industry is. Caveats apply, but monoliths can be built quickly and are easy to deploy, but they provide limited agility because even tiny changes require a full redeployment. We also know that monoliths often end up looking like a big ball of mud because of the way that software often evolves over time. For example, many monolithic systems are built using a layered architecture, and it's relatively easy for layered architectures to be abused (e.g. skipping "around" a service to call the repository/data access layer directly).
On the other side we have service-based architectures, where a software system is made up of many separately deployable services. Again, caveats apply but, if done well, service-based architectures buy you a lot of flexibility and agility because each service can be developed, tested, deployed, scaled, upgraded and rewritten separately, especially if the services are decoupled via asynchronous messaging. The downside is increased complexity because your software system now has many more moving parts than a monolith. As Robert says, the complexity is still there, you're just moving it somewhere else.
There is, of course, a mid-ground here. We can build monolithic systems that are made up of in-process components, each of which has an explicit well-defined interface and set of responsibilities. This is old-school component-based design that talks about high cohesion and low coupling, but I usually sense some hesitation when I talk about it. And this seems odd to me. Before I explain why, let me quote something from a blog post that I read earlier this morning about the rationale behind a team adopting a microservices approach. When we started building Karma, we decided to split the project into two main parts: the backend API, and the frontend application. The backend is responsible for handling orders from the store, usage accounting, user management, device management and so forth, while the frontend offers a dashboard for users which accesses this API. Along the way we noticed that if the whole backend API is monolithic it doesn't work very well because everything gets entangled.
The blog post also mentions scaling, versioning and multiple languages/frameworks as other reasons to choose microservices. Again, there are no silver bullets here, everything is a trade-off. Anyway, "everything getting entangled" is not a reason to switch from monoliths to microservices. If you're building a monolithic system and it's turning into a big ball of mud, perhaps you should consider whether you're taking enough care of your software architecture. Do you really understand what the core structural abstractions are in your software? Are their interfaces and responsibilities clear too? If not, why do you think moving to a microservices architecture will help? Sure, the physical separation of services will force you to not take some shortcuts, but you can achieve the same separation between components in a monolith. A little design thinking and an architecturally-evident coding style will help to achieve this without the baggage of going distributed.
Many of the teams I've spoken to are building monolithic systems and don't want to look at component-based design. The mid-ground seems to be a hard-sell. I ran a software architecture sketching workshop with a team earlier this year where we diagrammed one of their software systems. The diagram started as a strictly layered architecture (presentation, business services, data access) with all arrows pointing downwards and each layer only ever calling the layer directly beneath it. The code told a different story though and the eventual diagram didn't look so neat anymore. We discussed how adopting a package by component approach could fix some of these problems, but the response was, "meh, we like building software using layers".
It seems as if teams are jumping on microservices because they're sexy, but the design thinking and decomposition strategy required to create a good microservices architecture are the same as those needed to create a well structured monolith. If teams find it hard to create a well structured monolith, I don't rate their chances of creating a well structured microservices architecture. As Michael Feathers recently said, "There's a bit of overhead involved in implementing each microservice. If they ever become as easy to create as classes, people will have a freer hand to create trouble - hulking monoliths at a different scale.". I agree. A world of distributed big balls of mud worries me.