In my last post, I talked about how running spot instances for my flexible workload (pushing Logstash-consumed logs to Elasticsearch) saved me about 70% when compared to continuously running an on-demand EC2 instance. However, I’ve discovered a new way to tweak this cost savings even more. By utilizing AWS System Manager, to spin up an instance on a specific schedule, I now have an 83% cost savings over running an on-demand instance. Let’s breakdown the cost:
Just last year I became aware of serverless as a real thing: Running entire architectures without using a single server, and not having to maintain all the aspects of those servers. Instead, leveraging Cloud offerings to pick up the various functions of running applications in the cloud, spread across a potentially wide array of microservices. It wasn’t until I watched this video, from Serverless Conference 2017,that I became aware of how much of a game changer serverless technology is and will become. Spoiler alert: The gentlemen in this video serves hundreds of thousands of users per year and streams 4TB of content per day with a total serverless compute bill of just $700…still shocking to think about.
In my work environment, there are regulatory restrictions which currently prevent us from moving to serverless (or full cloud, for that matter), so I decided to tinker with it in my spare time. In fact, it’s where this website came from. Up until a few days ago, this website was built upon a serverless architecture which look something like this:
Mostly hosted on AWS, the individual components breakdown as Continue reading My *Serverless Website