Anyone else think the 50% cloud revenue increase might be due to spectre and meltdown mitigations? Those mitigations caused a 30% “slowdown” for many major workloads. That slowdown translates directly to CPU credits, i.e. minutes spent over baseline CPU usage. It also means any company operating servers with less than 30% headroom had to provision additional capacity to continue meeting demand.
AWS Meltdown patch only affected a very small portion of customers and the issue was fixed after rebooting the instances, so people who got affected were able to recover fastly, so no.
My understanding is the patch removed certain branch prediction optimizations. Systems post-patch were running some loads at 30% higher usage. I don’t see how a reboot would affect that?
It reduced capacity by 30% which means a 40% increase in cpu usage. Here is one of my teams RDS instances during the timeline of the patch: https://imgur.com/a/khGxU
Exactly. That’s what I was suggesting originally. It seems pretty cut and dry and would certainly explain the abnormally large increase in cloud revenue.
This is ongoing, correct? Since fundamentally the patch disables optimizations that would reduce cpu workload for many tasks?
