Stop asking: “Which EC2 instance should I use?” … and ask instead: “What does one unit of my work cost, at the latency my users need?”
That was the idea behind today’s EC2 performance workshop at SREcon EMEA ;-)
@iaasgeek.bsky.social
@iaasgeek.bsky.socialStop asking: “Which EC2 instance should I use?” … and ask instead: “What does one unit of my work cost, at the latency my users need?”
That was the idea behind today’s EC2 performance workshop at SREcon EMEA ;-)
@hightalk-me.bsky.socialAWS describes MoE kernel changes on SageMaker AI. HighTalk’s view: evaluate throughput and response latency against your actual workload.