Power Management
As required by a variety of workloads, ESXi is aggressively optimized by default to drive high I/O throughput efficiently by using fewer CPU cycles and preserving electricity. I/O latency must be reduced for many applications, though, even if it means using more CPU power and consuming more energy.
Several power management capabilities available in the hardware allow an ESXi host to fine-tune the trade-off between power consumption and performance. ESXi's use of these functionalities can be regulated by choosing a power management policy.
vSphere 5.0 and later by default use a "Balanced" power scheme, whereas earlier versions of ESXi used "High Performance" power schemes. The default "Balanced" power plan for crucial applications, including SQL Server, ought to be modified to "High Performance."
Memory Reservation
If attaining sufficient performance is your main objective, you should think about making the memory reserve equal to the memory that has been supplied. This will ensure that the virtual machine (VM) only receives physical memory and remove the chance of ballooning or swapping. Use the following calculations to determine how much RAM the virtual machine (VM) needs:
VM Memory = SQL Max Server Memory + ThreadStack + OS Mem + VM
Overhead
ThreadStack = SQL Max Worker Threads * ThreadStackSize
ThreadStackSize = 1MB on x86
= 2MB on x64
OS Mem: 1GB for every 4 CPU Cores
To find the maximum server memory size and worker thread count for SQL Server, consult your database administrator and utilize SQL Server memory performance metrics. For more on the VM overhead, see Table 4's VM overhead.
Reserving RAM could restrict vSphere vMotion. Only if the target ESXi host has unreserved RAM that is at least as large as the reservation may a virtual machine (VM) be moved.
When all memory is reserved, the swap file won't be created, which will free up disk space especially for virtual machines (VMs) with high memory allocations.
It is strongly advised to keep an eye on host swap-related counters (swap in/out, swapped) if the "Reserve all guest memory" checkbox is not selected. In the event that a host must use swapping as a last resort to assign physical memory to a virtual machine (VM), it is important to note that if congestion conditions are removed, the swapped VM memory will remain switched. The host will swap out some of the virtual memory if an ESXi host has memory congestion, say, during lengthy maintenance or in a disaster recovery scenario, and not all of the virtual memory is reserved. This memory capacity won't be automatically switched out. If the swapped memory is found, you can use the unswap command53, vMotion, or shut down and then restart a virtual machine.
0 Comments