Storage Virtualization

Virtual machine, virtual platform, virtual server and now storage goes vitual. Almost all storage vendors have some solution to offer on the same. 

Let's first understand what storage virtualization is. It is the pooling of various physical network storage devices so that they appear as one or more logical (virtual) volumes as if they were the actual devices. Applications see and work with these volumes without interacting with or knowing about the physical devices. These virtual volumes interact with the physical volumes. So, virtualization provides a layer of abstraction over the physical storage devices and this layer provides all the functionality that a storage device would otherwise provide.

Benefits of virtualization



Improved capacity utilization: With all storage devices pooled together, available storage can be better utilized and provisioned for increasing or decreasing capacity demand.

Faster data migration: In a heterogeneous environment, apps help migrate data from one storage device to another. The virtualization layer can perform this operation without data being moved, thereby quickening the process.

Automated capacity expansion: Policies are built into the virtualization engine to automatically allocate more

capacity to an app. The engine does this by utilizing free space from other devices. Or capacity is added physically to it.

Faster backup and better application testing: A virtualization engine sitting between the apps and storage can make fast point-in-time copies (snapshots) of the data without any app interference. So that instead of backing up live data, data of the snapshot can be backed up. The same snapshots can be used for testing development apps, which may damage the real data if tested on it.

Improved performance: If policies are set, the virtualization engine can move data that's in demand, to faster storage media such as fast disks or solid state storage.



High availability: Since the virtualization layer separates the apps from its physical storage, any physical damage to the storage or the servers, running the apps, will not affect the other.

Simpler volume management: As the physical storage components are hidden under the logical or virtual volumes,

volume management becomes easier and is associated with the logical volumes only.

Approaches to virtualization



There are three broad ways to achieve storage virtualization, depending upon where the virtualization engine resides in the storage network. The first approach is to put the engine on the storage device itself. It's achieved by the OS of the device or any other piece of software running over it and called appliance-based virtualization. In the second, array-based approach, the virtualization logic moves away from the storage devices and sits on a dedicated appliance. The appliance connects to the storage devices at one end and to the servers on the other. Third approach, called network-based virtualization is an approach recently introduced by EMC, in a conference held inSingapore. It's like the array-based approach, where the virtualization engine sits on a dedicated appliance. However, the network-based virtualization appliance doesn't directly sit between the servers and storage. It connects to the storage area network just like any other storage device or server would connect. But, by sitting parallel to the devices, on the network, an appliance cannot perform virtualization functions. Therefore, such an appliance works only in conjunction with special switches, which, unlike normal storage switches, can modify storage data packets, to provide virtualization. The switch gets instructions to operate on data packets from the virtualization appliance, thus, completing the partnership. EMC uses this approach in its Invista product line.

Approaches for virtualization provide almost similar functionality but differ in aspects such as performance, scale, management and cost. Both array and

appliance-based solutions, as they lie in the data path, introduce latency. While latency can be taken care of by caching, but caching may bring in data integrity and consistency issues. A network-based solution works with switches which can operate on the data at full-wire speed-avoiding latency. An

array-based or appliance-based solution can't scale as much as a network-based system, being directly associated with the storage. But, the network-based virtualization system needs special switches to be placed in the network. This increases the cost and requires network reconfiguration. So, the best suitable approach for you will depend on your current infrastructure, future requirements and, of course, money.

Anoop Mangla, inputs from EMC Conference, Singapore