What’s new in vSphere 6 Storage

In vSphere 6 storage will change substantially with the introduction of Virtual Volumes (vVols) and Virtual SAN 6. vSphere vVols enables your external storage arrays to become VM-aware. Storage Policy-Based Management (SPBM) allows common management across storage tiers and dynamic storage class of service automation. Together they enable exact combinations of data services (snapshots, clones, remote replication, deduplication, etc) to be instantiated more efficiently on a per virtual machine basis.

Virtual SAN

VMware Virtual SAN 6 has significant scalability and performance enhancements including new all-flash configuration support. VMware will also introduce the industry’s first solution to enable native Virtual Machine-awareness across a broad range of third-party storage systems, VMware vSphere Virtual Volumes (vVOLS). These two new launches are designed to enable mass adoption of software-defined storage.

VMware Virtual SAN 6 has several scalability and performance improvements over Virtual SAN 5.5: twice as many hosts per cluster (now at 64); twice as many VMs per host (now at 200); twice as many IOPS per host (now at 40K for the hybrid and 90K or four and half times for the all-flash) 16 times the snapshot depth per VM (now at 32 per VM); and 31 times the virtual disk size (now at 62TB). Virtual SAN 6 is an enterprise-class storage platform ideal for VM storage, is built for virtual infrastructure, and is ready for business-critical applications.

Virtual SAN 6 Highlights:

• New All-Flash architecture – VMware Virtual SAN 6 enables a two-tier all-flash architecture in which flash devices are intelligently used for both caching and data persistence. Tier 1 is used as 100% write buffer while Tier 2 is strictly for caching. This architecture provides more than four-times increase in IO throughput per node compared to VMware Virtual SAN 5.5 while delivering predictable sub-millisecond latency.
• Scalability increased to 64 nodes / cluster – As mentioned the new release doubles scalability to 64 nodes per cluster enabling customers to exceed 8 petabytes of storage capacity from a cluster.
• Maximum throughput of 7 million IOPS / cluster –A 64-node VMware Virtual SAN cluster now delivers up seven million input/output operations per second (IOPS) with nearly perfect linear scalability.
• New enterprise-grade snapshots – The release introduces a high performance and efficient snapshot capability increasing the snapshot depth to 32 per virtual machine while minimizing the performance overhead.
• New Rack-awareness – Intelligent placement of virtual machine objects across server racks for enhanced application availability even in case of complete rack failures.
• Expanded support for blades – With new support for direct-attached JBODs, customers can now scale VMware Virtual SAN 6 clusters to large capacity in server blade environments.

vSphere 6 Storage with vVols.

vVols will substantially change the way storage is architected and consumed. Traditional storage systems (external arrays without vVols) the LUN or volume is typically the unit of both capacity and policy. In other words, you create LUNs with fixed capacity and fixed data services. Then, VMs are assigned to LUNs based on their data service needs. This can result in problems when a LUN with a certain data service runs out of capacity, while other LUNs still have plenty of room to spare. The effect of this is that typically admins overprovision their storage arrays, just to be on the safe side.

With vVols, it is totally different. Each VM is assigned its own storage policy, and all VMs use storage from the same common pool. Storage architects need only provision for the total capacity of all VMs, without worrying about different buckets with different policies. Moreover, the policy of a VM can be changed, and this doesn’t require that it be moved to a different LUN.

Simply put, the goal of Virtual Volumes is to bring the benefits of the SDDC to storage. For the vSphere Admin, Virtual Volumes enables on-demand access to exactly the right kind of storage and storage services needed for applications, and for the storage admin it provides a more efficient way to provision and manage storage for vSphere environments.

How does vSphere 6 Storage with vVols work?

ESX manages the array through VASA (vSphere APIs for storage awareness) APIs and the array is logically partitioned into containers, called Storage Containers. Virtual machine disks, called Virtual Volumes, are stored natively on the Storage Containers. I/O from ESXi to the array is addressed through an access point called Protocol Endpoint (PE). Data services are then offloaded to the array.

The capacity, performance, availability, protection and security for an individual virtual machine on the underlying storage platform is managed through a policy on the vSphere level.

The vVol-architecture consists of the following components:

VASA Provider

The VASA Provider is a software component that is developed by storage array vendors and that exposes the storage services which a vVols array can provide. It also understands VASA APIs for operations such as the creation of virtual volume files. It can be thought of as the “control plane” element of Vvols.

Protocol Endpoints (PE)

Protocol Endpoints are the channel through which data is sent between the virtual machines and the arrays. They can be thought of as the “data plane” component of Vvols. PE’s are configured as a part of the physical storage fabric, and are accessed by standard storage protocols, such as iSCSI, NFS v3, and FC.

Storage Container

A storage container is a logical construct for grouping Virtual Volumes. It is set up by the storage admin, and the capacity of the container can be defined. As mentioned before, vVols allows you to separate capacity management from policy management. Containers provide the ability to isolate or partition storage according to whatever need or requirement you may have.

The main thing to note about storage containers is, a storage container maps to a vSphere Datastore. The concept of datastores is deeply embedded within vSphere and all associated products, and they are used for various purposes, such as administrative access control. Vvols provides a corresponding object so that you can continue to logically manage storage as you do with traditional arrays.

Storage Policy-Based Management framework (SPBM)

Instead of being based on static, per-LUN assignment, storage policies with vVols are managed through the Storage Policy-Based Management framework of vSphere. This framework uses the VASA APIs to query the storage array about what data services it offers, and then exposes them to vSphere as capabilities. These capabilities can then be grouped together into rules and rulesets, which are then assigned to VMs when they get deployed. When configuring the array, the storage admin can choose which capabilities to expose or not expose to vSphere.

This diagram above illustrates two views of vVols: one from the storage admin and one from the vSphere Admin. As you can see, vVols provides a way for a storage admin to understand and manage objects that vSphere admins have long used, such as datastores and virtual machine files. Storage policies, on the other hand, provide vSphere Admins with an easy way to consume capabilities that the storage admin has had to up until now keep hidden behind the construct of a “LUN”. It’s a win-win for everyone!