4 ms·
> Scality utilized two different models for testing. The read-intensive workload had 10% idle, 80% reading, and 10% writing. On the contrary, the write-intensiv
by cinntaile 3y ago
> Scality utilized two different models for testing. The read-intensive workload had 10% idle, 80% reading, and 10% writing. On the contrary, the write-intensive workload consisted of 10% idling, 10% reading, and 80% writing
These assumptions don't make sense when HDDs and SSDs have vastly different speeds so the actual percentages will be different. You can't just will them to be the same.
- photon_rancher 3y agoYup. Additionally, TB Capacity per Watt is meaningless except when idle or choosing PSU size. The relevant workload for energy efficiency has units of energy, not of power. And units of bits as well. They should be measuring bits/joule or TB/kWh for both r/w conditions, as that’s going to make the biggest difference. And you need to measure it at the system level because it’s more resource intensive to the CPU/chipset to support a higher bandwidth. I don’t think theres a realistic workload where the test ran makes sense (maxing out read writes per drive instead of normalizing for data rate).
- cinntaile 3y ago> And you need to measure it at the system level because it’s more resource intensive to the CPU/chipset to support a higher bandwidth. Good point, I hadn't even taken this into account.