Benchmark local PV

[wenhaocs]

Customer finds that when using local PV, the CPU utilization is abnormally higher than using EBS.

(1) Under the local pv cluster, 20 CPU of storaged is being used which is so high for just 2K QPS. Because prod (with EBS) is using 50 CPU for 25k QPS...
(2) The queries and dataset between two clusters are exactly the same.

Here is the configuration:

	Local PV cluster	Prod cluster
Storaged instance type	m6id.32xlarge 128cpu, 512 GiB Mem, 4 x 1900 NVMe SSD	c7a.48xlarge 192cpu, 384 GiB mem, EBS Only
# instances	18	18
# partitions	120

Dataset size: 2.5 TB

When benchmarking, please focus on the CPU utilization of storaged under the same QPS. Please use reasonable dataset size, so that dataset is not too small. Below shows the I/Os during user's test and we can see there are I/Os on disk.

[wenhaocs]

It's not required as the performance issue has been resolved.