13 521
правка
Изменения
Нет описания правки
Here’s an example setup from Micron. They used 2x replication, very costly CPUs (2x Xeon Gold per server), very fast network (100G) and 10x their best NVMes in each of 4 nodes: https://www.micron.com/resource-details/30c00464-e089-479c-8469-5ecb02cfe06f
They only got 350000 peak write iops with high parallelism with 100 % CPU load. It may seem a lot, but if you divide it by the number of NVMes — 350000/40 NVMe — it’s only 8750 iops per a an NVMe. If we account for 2 replicas and WAL we get 8750*2*2 = 35000 iops per drive. So… Ceph only squeezed 35000 iops out of a an NVMe '''that can deliver 260000 iops alone'''. That’s what Ceph’s overhead is.
Also there are no single-thread latency tests in that PDF. Such tests could be very interesting.
