Compute
VMWare’s hypervisor, ESXi, manages the hypervisors. As a type-1 hypervisor, ESXi is able to provide near-native performance to virtual machines. ESXi is installed as the hypervisor’s operating system, and will present a web-based management interface that enables managing the individual ESXi host. ESXi can run on just about any modern hardware.
Management
VMware vCenter is the management software that enables multiple hypervisors to be managed together. You will often hear a vCenter server referred to as a VCSA (vCenter Server Appliance). The VCSA takes ownership of the individual hypervisors and allows them to be configured together and share storage and networking. VMWare uses the name vSphere for the vCenter + ESXi platform. Cloud providers may also leverage an additional management tool, called VMWare Cloud Director to present VMWare infrastructure as a multi-tenant cloud platform ¸Í
Storage
Data Stores
VMware can mount a LUN or NFS share as a Data Store, and these Data Stores are where the virtual machines and their files are stored. As the same LUN or NFS export can be mounted by multiple hypervisors, these data stores can be shared. This allows VMs to run on any host in the cluster without having to worry about moving files. Data stores are formatted using a proprietary file system called VMFS.
VMware has a functionality called vSAN where multiple hypervisors can pool locally installed storage together to form a large virtual SAN.
Networking
Virtual Switches
A Virtual Switch, like a virtual machine, is a software representation of a physical switch. And, like a physical switch, a virtual switch is merely a collection of Virtual Switchports into which you can plug Virtual NICs. The Virtual Switch uplinks to Physical Switches via hypervisor NICs, and VLANs from the physical networks can be presented to the virtual switch for use on the virtual switchports. Virtual switches have 120 virtual switchports per host by default and have a maximum of 4096 virtual switchports per host.
Port Groups
A port group is a group of virtual switchports that will share the same settings. Most commonly, these settings define the VLAN(s) presented to your VMs, but there are, however, a few additional settings and features that available that may sound familiar if you’ve ever played with SPAN ports or port-security settings on a physical switch. These are:
- Promiscuous Mode – In normal switching frames are only forwarded to their destination MAC. In Promiscuous mode, every frame is forwarded to every device on the distributed port group. This behavior is intended to imitate SPAN ports on physical hardware but is also occasionally used by some HA/clustered devices.
- Forged Transmits – This setting allows a virtual machine to SEND traffic from a MAC address other than the one defined in the VMX file as belonging to a NIC.
- MAC address Changes – This allows a machine to ACCEPT traffic from a MAC address other than the one defined in the VMX file as belonging to a NIC.
You can define a port group as having a static number of switchports, or you can set it to grow dynamically as needed.
Distributed Virtual Switch
Having your hosts in a cluster managed by vCenter will allow you to create something called a Distributed Virtual Switch (DVS). A Distributed Virtual Switch allows you to manage the networking on every participating hypervisor/host by presenting the same virtual switch. A participating host must have physical uplinks to the same physical switching as the other participating hosts. Once a DVS is created, you can create Distributed Port Groups that allow VMs on different hosts to be on the same VLAN and talk to each other through the Distributed Virtual Switch This traffic leaves the uplink of one host and comes in the uplink of another; the distributed virtual switch is just an abstraction layer that simplifies the management of switching across the cluster.
Organization
Datacenters
Within vCenter, there are a few layers of organizational structure. The first is a Datacenter. A datacenter is a logical representation of a physical datacenter and everything in it: hypervisors (and the VMs on them), networking, and storage. Traditionally, a Datacenter in vCenter represents a physical location, although this is not necessarily required.
Clusters
Within a Datacenter, hosts are organized into clusters. A cluster defines multiple hypervisors pooled to share resources. Clustered hypervisors can participate in a Distributed Virtual Switch together and can share storage.
Resource Pools
Within a cluster, resources can be divided up into resource pools, constraining a group of VMs to be limited in the amount of CPU and RAM they can consume.
Folders
Finally, there exist folders. Folders are merely an organizational structure with no deeper meaning. They are just folders in which you sort VMs and storage. That’s all.
vApps
vApps allow a collection of VMs with tightly integrated purposes to be managed together, allowing for VMs to start and stop together in a preferred order with timings, among other things.
Permissions & Role-Based Access Control
VMWare vSphere can be set up with RBAC and single-sign-on, but it is designed to be single-tenant, meaning intended to be used by a single organization. while there are permissions and role-based access controls, there are limitations that make true multitenancy in a base deploy of VMWare vSphere impossible without additional pieces. VMWare is kind and gracious enough to provide those additional pieces, calling it VMWare Cloud Director.
Other Topics to Read About
Here are a few other key technologies used in Virtualization that won’t be required for this guide but are used heavily at Lightedge and elsewhere.
- vMotion allows machines to have their compute resources migrated from one hypervisor in a cluster to another without powering off the VM. You can also use Storage vMotion to migrate a VM from one LUN to another.
- DRS automatically vMotions machines to keep loads balanced across a cluster.
- Affinity & Anti-Affinity rules allow machines to be kept together for performance reasons or kept apart to eliminate the chance of a host failure causing an outage on both.
- High Availability allows vCenter to automatically restart VMs on a new host if the host they are on fails.
- Maintenance Mode is a way of taking a particular host out of service for patching or due to a suspected hardware issue. Putting a host in maintenance mode will start the process of vMotioning all VMs off it.