The DevOps behind successful on-line gaming companies are playing with George Washington’s axe. They need to know how to handle it.
The axe–familiar to those with a more classical education as Theseus’ ship–is the notorious one whose handle had been replaced three times, and head twice, but remained the same property. That’s how successful gaming hosts work: they grow and adjust so rapidly that over spans as short as a year, nearly all the infrastructure technology might be swapped out.
Customers don’t care, of course. Their stake isn’t in HTML5 vs. native coding. They don’t care that last year you traded Java for .Net, or even vice-versa, as part of beefing up for a larger audience. What matters to them, to the point that variations of it are a frequent topic of discussion, is “response time”. This phrase itself has a number of meanings, depending on where in the end-user experience (EUE) it’s applied; among them all, the one that most affects competitiveness is how fast the customer’s screen displays the result of a “play”, or customer action.
Conventional information technology (IT) generally styles this as “application performance”. A well-developed market in application performance management (APM) products promises to diagnose and even relieve problems with application response time. It’s an important job; make customers wait too long, and they quickly become former customers.
The difficulty with traditional APM and related tools is that they assumed relatively static and carefully-configured provisioning to accommodate a varying customer load. That’s simply not realistic: any alert DevOps team will at least experiment with application delivery networks (ADN), cloud hosting, crafted proxies and reverse proxies, specialized Web servers, and a variety of other technologies to keep up with growing usage. Such updates demand at least a whole new APM configuration from “first-generation” APM products, and some are simply incompatible with meaningful measurements.
As you begin to explore APM’s potential, therefore, make sure you start with tools that operate at the level of EUE, and autoconfigure details of internal implementation. By doing so, you’ll be able to continue to manage your operations even if your developers need to work in a different programming language, or your datacenter hypervisors rotate to a new licensing scheme.
On-line gaming appears poised for explosive growth in at least the US and UK over the next couple of years. Recognize that professional engineering to support that kind of growth requires not just “making all the parts faster”, but careful measurement of the end-to-end experience. APM makes gaming manageable, not just feasible. APM is one technology you’ll want to keep for the long haul, whatever other parts you swap and replace.
christine says
Hi! I’ve been following your blog for a long time now and finally got the bravery to go ahead and give you a shout out from Kingwood Tx! Just wanted to tell you keep up the good work!
Cameron Laird says
Welcome, christine. I need your help to “keep up the good work”: what’s your role in IT? Are you in operations, front-end design, system administration, …? “Real User Monitoring” ranges over a mix of topics and treatments, and I still have a lot to learn about what best serves readers. The idea from tomorrow’s piece, for instance, might save you *hours* of work just over the coming week. Does that interest you?