vSphere Performance - Storage Performance Basics

In the last of this trio of videos about performance, we look at Storage. You'll be familiar with most of this, but I encourage you to watch as you may fill in a few blanks. At the end of this lecture, you'll have an understanding of the 4 most common RAID levels and their impact on performance, and you'll know how many IOPS to expect from a given class of disk in a given RAID configuration.

Glossary for Module 8

vSphere Resource Scheduler
Not to be confused with the Distributed Resource Scheduler, the vSphere Resource Scheduler is responsible for scheduling time on the physical hardware for each VM to run.

CPU Ready
Describes the time a vCPU spends ready to run, waiting to be scheduled on the physical CPU.

Co-Stop
Describes the time a vCPU spends ready to run, but is waiting for the other vCPUs to also be ready.

NUMA
None Uniform Memory Access, describes how modern multi-processor systems group memory and CPU in to nodes, with each CPU having a direct link to a subset of RAM modules.

vNUMA
A VMware technology that advertises NUMA nodes to a VM to reflect the NUMA nodes in the physical system, to allow the VM to optimise performance.

IOPS
IO Per Second - a crucial measure of storage performance

RAID
Redundant Array of Inexpensive Disks - a mainstay in server storage, used in both servers and workstations for many years to provide resiliency and performance for storage systems

Memory Balloon
The first sign of an issue with memory contention, memory balloon is a vSphere feature used to reclaim unused RAM from virtual machines when physical RAM comes under constraint.

Module Summary
The topics covered in this module should be seen as a bonus, to provide a more in-depth insight in to the inner workings of vSphere. We’ll cover these in more depth in the Professional level course, but look for these common mistakes in your own environments in the meantime.

For reference:

7200 RPM disks can provide 80 random IOPS
10K RPM disks provide 130 random IOPS
15K RPM disks provide 180 random IOPS
SSDs can provide 1000+ random IOPS
The latest NVMe SSDs can provide 100,000+ IOPS

Also, for the different RAID levels:

RAID 0 has a write penalty of 1
RAID 1 has a write penalty of 2
RAID 5 has a write penalty of 4
RAID 6 has a write penalty of 6

Finally, remember that CPU Ready shows the amount of time a vCPU spends Ready to run. There a number of reasons this might happen, and one of them is co-stop.

References
https://kb.vmware.com/s/article/2001003
https://kb.vmware.com/s/article/1005362

Видео vSphere Performance - Storage Performance Basics канала E-Learn IT