The concept of software defined storage (SDS) is an integral facet of a modern software defined data centre. As a critical part of this architecture, SDS supports organisations’ transition to data-driven, highly digital business models.
“As the volume of data that must be stored, analysed and retrieved continues to grow exponentially, SDS is a key enabler of this business transformation,” says Anton Jacobsz, MD of Networks Unlimited.
With this in mind, he explains, many CIOs are recognising that hyper-converged virtual storage area networks are one of the most effective ways to achieve true and sustainable SDS.
Traditionally, storage has presented challenges for IT teams, considering:
- Deciding where to even house the virtual machine (VM) in the first place presents challenges – as administrators must ensure the logic unit numbers or volumes that are presented as datastore for the VMs have the right storage capabilities.
- The initial setup needed to install, deploy, configure, manage and monitor storage can become very complex.
- With a number of virtual machines (each with many different applications) all residing on the same LUN or volume on the storage array, the problem arises that each application has different I/O profiles, block sizes, read/write ratios, and so on.
- Known as the I/O blender effect, the result is a lack of end-to-end visibility into VM performance from the end-user to the back-end storage.
- In designing one’s storage strategy, administrators must carefully balance the need to meet performance requirements of application owners, while still providing an adequate Recovery Point Objective (RPO) value.
Policy-driven approach
Jacobsz makes reference to a white paper published by Rubrik[1], a Cloud Data Management company named by Gartner as a Visionary in its 2017 Magic Quadrant for Data Centre Backup & Recovery Solutions, and a brand distributed by Networks Unlimited in sub-Saharan Africa. The paper details VMware’s SDS vision and how Rubrik builds on that vision by wrapping data in an intelligent software fabric for automated backup, recovery, archival, search, cloud, and development.
“By adopting the SDS principles of abstracting and pooling, the administrator’s task can be dramatically simplified,” says Jacobsz, “so they no longer need to specify how the VM should be provisioned to ensure it’s in the appropriate location every time.”
To achieve this, the underlying storage and data protection fabric is abstracted away from the administrator, using policy-based management.
In fact, this policy-driven approach is at the heart of VMWare’s software defined storage vision, with its Virtual Storage Area Network (VSAN) being hyper-converged in nature. Now, when VMs are provisioned, the administrator doesn’t need to manually determine which LUN or volume to place it on.
This datastore can then be used for provisioning VMs – which are deployed onto storage based on the policy associated with the VM.
Introducing VSAN
VSAN is completely integrated into the vSphere hypervisor (thus, no need for additional appliances, VIBs, patches or plugins). A vSphere administrator need only enable the functionality to get started.
Ongoing management and monitoring of VSAN is made simple via the vSphere Web Client, giving administrators a clear view of how much space the constituent parts of a VMs deployed on a VSAN datastore are consuming, as well as seeing performance all the way from virtual machine to the physical storage devices.
VSAN also includes a comprehensive health check system, continuously monitoring all the different components of a VSAN system, from virtual machine objects to networking to physical disks.
“In the modern era, next-gen data centres are focusing on simplicity, scale, security, and automation,” he notes, adding that the VSAN solution helps to create more frictionless workflows for administrators and reduce pressure on the technical teams.