4 ms·
On AWS you can rent ".metal" instances which are probably more stable for benchmarking. I tried this once for fun on a1.metal because I wanted access to all har
by Matumio 5y ago
On AWS you can rent ".metal" instances which are probably more stable for benchmarking. I tried this once for fun on a1.metal because I wanted access to all hardware performance counters. For that it worked. My computation was also running slightly faster (something around 5% IIRC). But of course you'll have to pay for all its cores and memory while you use it.
- throwdbaaway 5y agoYeah, that's exactly what the GCP engineer recommends, and likely why the final benchmark in the article was done using a c5n.9xlarge. Still, there is no guarantee that after stopping the instance on Friday evening, you would get back the same physical host on Monday morning. So, while using dedicated hardware does avoid the noisy neighbor problem, the "silicon lottery" problem remains. And so far, the data that I gathered indicates that the latter is the more likely cause, i.e. a "fast" virtual machine would remain fast indefinitely, while a "slow" virtual machine would remain slow indefinitely, despite both relying on a bunch of shared resources.