VMworld San Francisco starts this Sunday, and VMware’s new VVol (Virtual Volume) technology is one of the most eagerly anticipated topics being discussed by Nimble Storage’s customers, partners, and product teams. In Part 1 of this series, we briefly discussed the high-level concepts and benefits of Virtual Volumes. In this post, we will dive deeper into the architecture of VVols, and how Nimble’s implementation allows one to take full advantage of policy-based management for storage. We will wrap up with a sneak preview of what we plan to show of VVols next week at the show.
VASA Provider (Management Plane)
VMware’s VASA (vStorage API for Storage Awareness) is a management framework that allows deeper communication between a vCenter Server and a storage array. Without VASA, vCenter Server has no visibility into the storage array, and therefore the storage array and the volumes provisioned to the ESX cluster hosts, appear as a black box. With the VASA framework, the road to policy-based management is paved – the storage array can now advertise all its capabilities to vCenter Server, thereby allowing vSphere / storage admins to define VM storage policies based on capabilities offered by the storage array. Given the importance of VASA Provider, this service needs to be highly available.
Protocol Endpoint (Data Plane)
Protocol Endpoint (PE) is a special logical unit that is provisioned to the ESX host, serving as the core demultiplexer for IO to the virtual volumes. With a Protocol Endpoint configured on each ESX host, the path selection and failover are handled in one place, as opposed to each volume, as was previously the case. The Nimble multipathing integration (Nimble_PSP_Directed plugin) works in conjunction with the Protocol Endpoint to provide I/O access to all VVols used by the virtual machines.
The notion of VMFS volumes / datastores no longer exists in the world of VVols. Instead, a new abstraction layer is created – a VVol container. A VVol container is typically associated with a capacity and a set of capabilities. Consider it a pool of storage resources for virtual machines to consume. Virtual volumes are provisioned on the VVol container / datastore, based on the policies selected during virtual machine creation. VASA management plane will perform the compatibility check to ensure the policies are in compliance with the underlying VVol container.
VM Storage Policy
This is the most visible end-user facing component of VVols – the heart of VMware’s SPBM (Storage Policy Based Management). After the storage capabilities are discovered by the vCenter Server, policies can be created based on application performance, data protection, business continuity, and security requirements. For example, sensitive production applications may require regular backup at frequent intervals, with Encryption at Rest for all virtual disks. One can easily create a “High Security Production” policy, and select all relevant capabilities offered by the storage array. For example, an application might require an hourly snapshot, with replication to DR datacenter, and Encryption at Rest. Once the policy is created, it can be selected by the application owner / vSphere admin upon creation of the VM. The VASA management plane will perform compatibility checks against the underlying VVol container, and only present compatible container(s) for placement if all capabilities from the policy are satisfied. With this approach, storage provisioning for virtual machines is entirely policy driven, thereby keeping things simple and consistent.
Nimble Storage’s Implementation
In the previous post, we also mentioned that not all VVol implementations are created equal. So, what’s different about Nimble’s implementation?
Our Adaptive Flash platform is advantageous across multiple dimensions, including simplicity, scalability, efficiency, high availability, high performance and built-in data protection. We’ve followed all of the same principles in designing and building our VVol implementation. We start with simplicity and high availability – the Nimble VASA Provider actually lives inside the Nimble array. This approach eliminates the need to install the provider service in a dedicated VM / physical machine, and ensures high availability for this service. End users simply add a VASA Provider by supplying the URL of the service running inside the Nimble array, without having to worry about binary download / OS compatibility / high availability design for this service. In terms of efficiency and high performance, the Adaptive Flash platform has space-efficient redirect-on-write snapshots with inline compression from day one, allowing customers to store snapshots from 30-60 days on average, and replicate compressed snapshots for disaster recovery. Zero Copy Clone technology allows instantaneous creation of volume clones with no data duplication. Nimble’s integration with VVol allows vSphere to fully take advantage of these capabilities with a simple policy association: CASL variable block size filesystem, snapshot, clone and replication offload to the array, on a per-virtual-volume basis.
If you’re interested in trying out our VVol integration at VMworld, come by the Nimble Storage booth (#1521) and take the VVol hands-on lab! We will host our VVol hands-on lab in the cloud, for easy access in any available demo station, or even your personal laptop! Five modules are available to get you up and running with Virtual Volumes:
You can also learn more about VVols at the conference session on Software Defined Storage: Satisfy the Requirements of Your Application at the Granularity of a Virtual Disk with Virtual Volumes (VVols), taking place Tuesday, Aug 26, from 2:00 to 3:00 PM (session #STO3162).
Plus we’ll be presenting a “VMware VVols Integration Technical Deep Dive” webinar on September 23rd at 11am Pacific; you can register here.
Here’s a link to Part 1 of the VVol blog series.
- Wendel Yu