VMware vsphere 5.1 cluster Walkthrough (12) DRS dynamic quotas

Part II vsphere DRS (distributed resource Scheduling)

Chapter III the DRS dynamic quotas

In this section, we will explain the concept of DRS dynamic quotas and resource management. Understanding dynamic quotas and resource pools, as well as resource allocation settings, can make it easier for you to troubleshoot DRS behavior and achieve optimal performance for your virtual machines.

Before delving into DRS and local host resource management, we need to step back and master the dynamic resource quotas.

Dynamic quotas define the available target for the ideal resource, DRS, and locally invoked, depending on the virtual machine or available or unavailable resources, quotas contain static elements and dynamic elements, static elements are based on user-supplied resource specifications, dynamic factors are based on estimating requirements and system contention levels, Each resource type (CPU and memory) of a virtual machine will have a separate dynamic quota target.

As an administrator, you can affect the quota of dynamic virtual machines by setting resource allocation policies (retention, shares, and restrictions). Resource allocation settings affect not only the performance of virtual machines, but also the performance of other virtual machines, so it is important to understand how dynamic quotas are calculated and how virtual machines are configured without introducing denial of service or other environments.

Whether it's dynamic or static, this chapter is explained in detail later, so let's start by designing the schema for calculating dynamic shares.

Resource Scheduling architecture

The Vmkernel of the ESXi host runs multiple local resource dispatchers, including CPU scheduling and memory scheduling. DRS introduces global scheduling, effectively creating a two-tier schedule for allocating local and cluster resources.

Figure 51:drs and local scheduling

DRS scheduling

Global scheduling is responsible for allocating cluster resources, and DRS determines the dynamic quotas for each virtual machine when it receives the requirements of valid resources and virtual machines.

If the cluster is a large host, but relies on host-level scheduling to implement the resource settings for the DRS resource pool and virtual machines, DRS scheduling calculates the ideal CPU and memory quotas, and the resource pool expands on the 13th chapter.

Interestingly, when the resource pool contains virtual machines running on different hosts, the local host's resource scheduler allocates resources to the virtual machines, requiring conversion between the cluster resource pool settings and the settings of the local host resource pool.

DRS resolves the resource pool tree for each host cluster by mirroring, mapping the appropriate resources to each resource pool node, and local resource scheduling in the/host/user directory, causing the DRS resource Chishu to restart/host/user on each host that contains the resource pool.

Figure 52: Mapping the cluster RP tree to ESXi host RP Tree