vSAN is the VMware Hypervisor-Converged Software (HCS) storage solution, which comes fully integrated into the VMware ESXi™ kernel, unlike many of the other third-party options, which typically employ a storage virtual machine that runs alongside the other workloads to provide the storage platform. As a result, vSAN allows storage and compute to scale together, like building blocks, allowing for virtual machines to be consolidated into a single host device, with storage provided within the hypervisor itself.

This is achieved as vSAN aggregates locally attached disks within the VMware vSphere® cluster to create a storage solution through a distributed datastore, available exclusively to the cluster, which can be employed to provision virtual machines from VMware vCenter Server®, in the same way that vSphere administrators are accustomed to.

Unlike the traditional VMware vSphere VMFS datastore, vSAN provides an object-based storage datastore model, in that vSAN stores and manages data in the form of flexible data containers, referred to as objects, to provide virtual machine-centric storage services, and deliver capabilities through a Storage Policy-Based Management (SPBM) platform. The SPBM platform and virtual machine storage policies are designed to simplify virtual machine storage placement decisions for vSphere and vCloud administrators. The goal of SPBM is to provide both high availability and scale-out storage capabilities, in the context of quality of service (QoS). This is achieved through the creation of virtual machine storage policies, which are employed to define the level of availability, and in some cases performance, required on a per-virtual machine or virtual disk basis. In addition, vSAN is fully integrated with core vSphere features such as VMware vSphere vMotion®, VMware vSphere High Availability, and VMware vSphere Distributed Resource Scheduler™ (DRS).

A vSAN distributed datastore can be constructed with as few as three vSphere ESXi hosts, although for most use cases four hosts are recommended, each containing at least one disk group with at least one SSD flash drive and one mechanical drive, as illustrated in Figure 1. vSAN can also support up to seven magnetic drives per disk group, and up to five disk groups per host, or up to 35 drives per node, whichever maximum is reached first. In the vSAN hybrid model, the VMware virtual machine files are stored on the mechanical drives, with the SSD flash drive providing read caching and write buffering. Therefore, the SSD drives in each disk group are not used for capacity, but are employed for cache and buffering only, with 70 percent utilized for read cache, and 30 percent for write cache. All writes are targeted to the SSD first, before being destaged to the capacity mechanical drives. vSAN has the requirement that all of the drives are presented individually to the host. Therefore, a storage controller is required that supports JBOD or pass-through mode, so that the ESXi host can see each disk individually. vSAN 6.0 supports scalability of up to 64 nodes per cluster, 200 virtual machines per host, and up to 6400 virtual machines per vSAN cluster.

 

From the release of vSAN 6.0, the platform also provides support for an all-flash architecture, on flash devices that deliver high, predictable performance and low latency, with response times of less than a millisecond for some of the most storage-intensive applications. However, do not assume that an all-flash configuration always provides higher performance than a hybrid solution. The actual performance is based on a number of factors including, but not limited to, disk group design, active dataset, workload read/write profile, and storage hardware, including disk storage controller performance capabilities.

Note that for simplicity, this document focuses on the hybrid disk group solution, and not an all-flash configuration. However, the concepts discussed in this paper are equally pertinent to both vSAN models, hybrid and all-flash.

 

The capacity of the vSAN datastore is dictated by the number of mechanical disks per disk group in the host, and by the number of hosts in the cluster. vSAN is a scale-out solution, in which more capacity and performance can be achieved by adding more disks to a disk group, more disk groups to a host, and more hosts to the cluster.

In a vSAN design, the SPBM platform plays a major part in defining the way in which vCloud administrators can use virtual machine storage policies to specify a set of required storage capabilities for a specific workload, based on the requirements for the application running in the virtual machine.

The following vSAN datastore capabilities are available in vCenter Server, configurable under VM storage policy:

 

vSAN leverages a RAIN (Redundant Array of Independent Nodes) mechanism and employs RAID for its object mirroring capabilities, and also what is referred to as Erasure Coding, which provides data striping in an all-flash vSAN configuration. By using parameters defined in the storage policies, administrators can assign each virtual machine to withstand one or more disk failures, or one or more host failures, providing redundancy and uptime. In addition, the storage policy also allows administrators to define the number of mechanical disks that the virtual machine resides across, by striping the object across multiple disks, potentially increasing performance for data that is not currently residing in the read cache. For more information on defining the these capabilities, find the VMware vSAN documentation pages at https://www.vmware.com/products/virtual-san.

The key design factors associated with the design and implementation of vSAN include the following: