Sergey Nivens - Fotolia

Get started Bring yourself up to speed with our introductory content.

Use storage system features to granularly manage virtual machines

Arun Taneja explains the importance of managing storage granularly and which vendors are meeting the demand for VM-level visibility.

To understand how to manage virtual machines in a virtual server environment, it is helpful to examine the situation from a storage perspective.

Consider this scenario: There are 20 virtual machines (VMs) running in a physical server (or in a number of physical servers), and they are all being serviced by a SAN. In this classic example, storage managers would create a LUN and allocate a certain amount of storage from that LUN to each of the 20 VMs.

At this point, the storage knows information only at the LUN level. But what storage administrators really want is information about the virtual machines at the VM level.

Here is another scenario to consider. You want to replicate a particular VM that handles a mission-critical application to two different places across a wide area network. But you also have a variety of other VMs running in the same cluster that are not important and may not need any replication. However, if you can replicate only at the LUN level, the only way to manage that storage is to replicate the entire LUN to replicate that one critical VM.

In this example, a storage administrator wants a storage system smart enough to know that VM No. 1 is a priority -- and that it must deliver a certain quality of service to that VM and restrict others if need be.

There are now vendors with products designed specifically to address this problem, and Tintri was the first company to enable storage to be managed at the VM level. If you ask for particular parameters at the VM level, you get all those things even though there are storage-related parameters. A majority of storage companies and storage products in the industry still do not provide this level of granular, VM-level manageability.

In today's landscape, VM-level visibility is the so-called Holy Grail. Every storage vendor aspires to deliver on this, but it is hard for traditional storage architectures to add VM-level manageability because it requires a fundamental shift. It is an easier task for new vendors to build in this capability, but many do not have this feature. These vendors, along with traditional vendors, will now rely on VMware Virtual Volumes (VVOLs) to enable this functionality. Of the new players who provided this capability from the start, Tintri is a prime example, but there are others. For example, NexGen built quality of service at a VM level, which provided valuable storage information at that level. Since the acquisition of NexGen by Fusion-io, however, there has been little news about this product. Other vendors that can provide this capability are Gridstore, Maxta and VMware's Virtual SAN.

Storage managers should keep in mind that all hyper-converged system vendors deliver this functionality, since they are all VM-centric to start with. For example, Nutanix, Scale Computing and SimpliVity all provide data (storage and non-storage) with VM-level granularity. Down the road, all players will deliver on this via VVOLs (which become available sometime in 2015). The amount of visibility and manageability that VM-level functionality brings to an IT admin is so powerful that any storage vendor without this capability by the end of 2015 will likely suffer.

Next Steps

Will VMware VVOLs affect you?

Limitations, advantages and choices presented by hyperconverged systems

The VM management system and capacity planning

Dig Deeper on Storage for virtual environments

Start the conversation

Send me notifications when other members comment.

Please create a username to comment.