The idea of auto-scaling, geo-redundancy, and high-availability for internet applications has been around for a number of years, however, many of the traditional scaling and failover patterns overlook regional outages and cascading failures. The most common practices have applications scale as demand increase or in the case of regional outages.

If you’ve ever wondered if there is an alternative way to manage these risks, this talk is for you!

We’ll look at the concept of constant work as an alternative to scaling on-demand and avoiding cascading failures when systems fail. Particularly, we’ll dig into systems that implement this idea of constant work, how to deal with the trade offs of scale and cost, and where may be good areas for you to invest in this idea.

Lastly, we’ll dig into a concrete example by showing off a system with both methods implemented to test our theory in practice!

https://www.youtube.com/watch?v=tR0zpjbTKoI