I follow a certain philosophy when developing system architectures. I assume that very few systems will ever exist in a consistent form for more than a short period of time. What constitutes a “short period of time” differs depending on the specifics of each system, but in an effort to quantify it, I generally find that it falls somewhere between a week and a month.
The driving forces behind the need for an ever changing architecture are largely business requirement based. This is a side effect of the reality that software development, in most cases, is used as a supporting role within the business unit it serves. As business requirements (i.e. additional features, new products, etc.) pour forth, it is the developer’s job to evolve their software system to accommodate these requirements and provide a software based solution to whatever problems lay ahead.
Given that many businesses can be identified as having the above characteristics, I can now begin to explain why I believe that Heterogeneous System Architectures hold a significant advantage over Homogeneous System Architectures, in many distributed system cases.
An Experiential Use Case
I work daily on a mobile platform that mixes web applications, server applications, and XMPP servers together into an architecture that delivers on the business requirements, as set forth by our business development team and clients. This mobile platform is fairly distributed, with web applications on one set of machines, server applications on another, and XMPP servers on yet another. These three components interconnect, with the XMPP protocol and/or a database being the glue that binds them.
It’s important to point out that each individual web application and server application interfaces with XMPP or a database only, and that each is independent of the other – we’ve essentially developed a “shared-nothing” architecture. This loosely coupled, non-interdependent XMPP and database architecture has served our purposes well so far.
We currently run on the Windows platform and utilize the .NET Framework to build both our web applications and our server applications. The only exception to using Microsoft technologies within our platform is the usage of a Java based XMPP server (Jive Software’s OpenFire).
When we were researching XMPP servers at the time, we found that there were really no exceptional .NET based XMPP servers with the support and maturity we needed. This lead us to our currently implemented Java based XMPP server solution. Given the nature of XMPP, and the openness of its protocol, we were able to find a couple solid .NET based XMPP client SDK’s with which to work with, that made interfacing with the Java XMPP servers trivial.
Just to drive my previous points home: turning to a Heterogeneous System Architecture wasn’t intentional; it was a side-effect of working with the 3rd-party software available at the time. This decision had essentially turned our all-Microsoft based Homogeneous System into a Heterogeneous system overnight.
One added benefit of this shared-nothing architecture, is that because it is based on an open protocol (XMPP) and uses databases with wide support throughout a variety of platforms and development languages, we are now able to mix and match platforms, differing technology stacks (LAMP), and development environments with minimal effort and with near zero incompatibility issues.
Differing Platforms, Technology Stacks, and Development Environments
Having available to us the ability to work with differing platforms, technology stacks, and development environments, all in an effort to find the best tool for the job (based on budgetary requirements, different developer skill sets, the availability of certain 3rd-party software, etc), gives us a real advantage in that our options have been vastly diversified.
Now, of course, the desire to utilize any solution should never be dependent solely on its availability as an option. Just because we can mix and match, doesn’t mean that it is within our best interest to do so. Our decision to derive value from this ability is highly dependent on a number of factors. For example, we’re a startup division within a larger company, our roots are startup based, and so is our budget. In going over some of this year’s software purchases, we’ve had to include Windows Server purchases/upgrades, SQL Server licenses, and Visual Studio development environment licenses. Those are just a few examples of common large ticket items that need to be considered when working with a startup’s budgetary concerns.
The flipside to the above software purchases are that our platform, although currently based on a majority of Microsoft software, is in no way dependent on that software existing at its current majority share. We could, for example, move to MySQL and eliminate the high cost of SQL Server licences, or we could move to a LAMP stack and decrease the costs associated with running our web servers on Windows Server. And again with the LAMP stack, we could opt to run a mature PHP based MVC framework instead of working with the ASP.NET MVC Framework which requires us to work through various Beta/RC1 bugs.
The above is just one budgetary example of where having the option to become more heterogeneous with our architecture may be in our best interest. Obviously, using one set of tools versus another has its trade-offs, and no one platform, technology stack, or development environment is the best in all situations. But, given how dynamic our business requirements have been over the last few years, I’m excited to say that we have a whole slew of options available to us, largely as a result of having a Heterogeneous System Architecture.
Heterogeneous System Impact on Human Resources
I strongly believe that the single greatest detractor from using a Heterogeneous approach to your system’s architecture is the human resource factor. When architecting a system, it is important to keep in mind the skill sets of your developers, the quantity of those developers, and the ability for those developers to work within a Heterogeneous environment.
It is far more common than not, to have a team of developers aligned with a particular technology. For example, a Microsoft system often has employed developers that follow the Microsoft path of technologies (.NET, SQL Server, Windows). Whereas with an Open Source system, it is more common to have a crew of developers that are well versed in working with Linux/Unix based technologies (Red Hat Linux, Apache, MySQL, PHP). The occurrence of this type of specialization amongst developers makes having a Heterogeneous Architecture somewhat harder to hire for.
Along these same lines, another unfortunate commonality between developers aligned with a particular platform is that they are often reluctant to learn and evolve their skill set outside of their realm of specialization. Finding the ideal type of developer that simply loves all technology and has the capacity to apply their general development knowledge across many platforms, quickly and successfully, is a rare breed indeed. Understandably so, as it takes a very dedicated and smart individual to become specialized in more than one platform. Equally as difficult as finding the will within developers, is finding developers with more than an introductory level of experience with multiple platforms.
So, in cases where your system is going the Heterogeneous route, ideally you would try to hire the particularly “smart” technology passionate developers; developers who have no allegiance to any particular technology stack. These are the guys who just want to use the best tool for the job, and have fun doing it; these guys are the Rockstars of the development world. Now, it’s easy to say “just hire the Rockstars”, but this usually comes at an increased salary cost and it is therefore largely more difficult to fill positions of this type. The question then becomes, based on your particular business requirements, your current team, and the general financial outlook of your business, whether it is to your benefit to go with a Heterogeneous, and therefore extremely flexible, system, or a more easily manageable and more thoroughly supported, Homogeneous system.
Getting personal for a moment, I prefer the flexibility of a Heterogeneous system, especially when it comes to working within a distributed architecture. The freedom and sheer number of options available to solve the often more complex software problems associated with distributed systems, makes it worth the extra high developer requirements.
I attempt to weave into my posts, the common theme that all implementations involve trade-offs. I try to drive home the reality that there is no single “best” way to accomplish any sufficiently complex task. That every strategy that you implement, and every decision you make has consequences in addition to its benefits, is one of the more important fundamentals of building robust system architectures.
I also try to illustrate these concepts using real life examples and I hope that they help you to better relate their applicability to your situation, however different it may be. Please comment below on what your experiences have been with Heterogeneous versus Homogeneous Architectures. I’m always interested in hearing how others tackle this subject, and what rules they’ve placed as guidelines for the implementations of their systems, one way or the other.