The scale and dynamic nature of vVols intrinsically changes VMware storage provisioning. To provide scale and flexibility for vVols, VMware adopted the T10 administrative logical unit (ALU) standard, which it calls protocol endpoint (PE). vVols are connected to VMs through PEs acting as subsidiary logical units (SLUs, also called sub-luns).
The FlashArray vVol implementation makes PEs nearly transparent. Array administrators seldom deal with PEs, and not at all during day-to-day operations.
A typical VM has multiple virtual disks, each instantiated as a volume on the array and addressed by a LUN, the ESXi Version 6.5 support limits of 512 SCSI devices (LUNs) per host and 2,000 logical paths to them can easily be exceeded by even a modest number of VMs.
Moreover, each time a new volume is created or an existing one is resized, VMware must rescan its I/O interconnects to discover the change. In large environments, rescans are time-consuming; rescanning each time the virtual disk configuration changes is generally considered unacceptable.
VMware uses PEs to eliminate these problems. A PE is a volume of zero capacity with a special setting in its Vital Product Data (VPD) page that ESXi detects during a SCSI inquiry. It effectively serves as a mount point for vVols. It is the only FlashArray volume that must be manually connected to hosts to use vVols.
Fun fact: Protocol endpoints were formerly called I/O de-multiplexers. PE is a much better name.
When an ESXi host requests access to a vVol (for example, when a VM is powered on), the array binds the vVol to it. Binding is synonym for sub-lun connection. For example, if a PE uses LUN 255, a vVol bound to it would be addressed as LUN 255:1. The section titled vVol Binding describes vVol binding in more detail.
PEs greatly extend the number of vVols that can be connected to an ESXi cluster; each PE can have up to 16,383 vVols per host bound to it simultaneously. Moreover, a new binding does not require a complete I/O rescan. Instead, ESXi issues a REPORT_LUNS SCSI command with SELECT REPORT to the PE to which the sub-lun is bound. The PE returns a list of sub-lun IDs for the vVols bound to that host. In large clusters, REPORT_LUNS is significantly faster than a full I/O rescan because it is more precisely targeted.