Breaking down the monolith
One of the innovative features of our flagship product, Flametree Attribution, is that it has a Web API. In this post I consider why this delivers for our clients.
In the 1990s, you used to be able to buy combined TVs, VCR and DVD players in one unit. This was great - until a new format came along, or one component broke down.
Today, consumer electronics has moved back to offering separate, interoperable components, augmented or replaced by remote services. The TV in your living room does one thing – displaying media – but can play DVDs from an external player, and talks to Netflix and other services via the Internet.
As well as making things simpler for the user, this approach also follows sound engineering principles. Consider the guiding principles behind Unix, articulated 25 years ago by Peter H Salus:
- Write programs that do one thing and do it well.
- Write programs to work together.
To this list we’d add a third that wasn’t available back then:
- The program can be anywhere.
All this has become possible due to a convergence of Web development frameworks such as Django and Flask, abundant computer power, and common communication protocols between clients.
Builders of such systems use a standard vocabulary of terms to describe what they’re doing: idempotent, RESTful, JSON. The details aren’t important; what matters is that there are standard ways to hook such systems together, typically using ‘glue’ code such as Python, a widely used programming language.
Instead of having to install software locally – which requires resources, permissions and the like – you can now run our attribution engine remotely, using nothing more than a few lines of code and some login credentials. We take care of the hardware requirements transparently, including specialized hardware to run at much higher speeds than are typically available on OEM production servers. Since some of our users have portfolio counts in the thousands, we also provide a filestore and a queue, allowing long jobs to be batched.
To put it another way, a Flametree user now doesn’t have to know anything about how the program works or where it runs. The language she uses, or the computer she runs, is now irrelevant, whether it be a Linux box or a Windows server. We’ve even run Flametree on a $20 Raspberry Pi. There is no need for certification or lengthy installation processes. Just install Python and you’re off!
Contact us to find out more.
