Kyle Somers’ Post

View profile for Kyle Somers, graphic

Sr. Manager, Solutions Architecture - Gaming Industry at Amazon Web Services (AWS)

Within the AWS community I've observed the existence of a super secret club of customers; if you're operating at such a scale that you needed to shard traffic across load balancers, then you're in the club. For everyone else, give this blog a read as it thoroughly describes an important topic on resiliency. https://lnkd.in/gVwnqnkj

Scaling strategies for Elastic Load Balancing | Amazon Web Services

Scaling strategies for Elastic Load Balancing | Amazon Web Services

aws.amazon.com

Lewis Carroll

Customer Success Engineering Leader

6mo

If you have to think about load balancing you should also be thinking about how “hot” you run your EC2 instances. With EC2 now having a mix of SMT* and non-SMT types, this matters. Get it wrong and your compute idle time goes up. Get it right and your cost to serve goes down. Non-SMT instances like Graviton and C7a/M7a/R7a scale more linearly with CPU utilization and degrade more gracefully at high utilization. Push that utilization threshold into the 80%+ range. SMT is Simultaneous MultiThreading, AKA HyperThreading

To view or add a comment, sign in

Explore topics