Guide to Storage Virtualization
Five Buying tips for buying storage virtualization products
By Deni Connor, Network World 10/1/07
Storage virtualization – the logical abstraction of data from its physical devices -- is a platform that enables a lot of different storage services. By virtualizing storage into a common resource pool, companies can receive better management capability, higher utilization of storage and improved migration, replication and storage provisioning capabilities. Despite its advantages, however, buying storage virtualization products can be costly and complex to implement.
Once you've decided that storage virtualization is indeed right for your network, consider these tips for buying virtualization gear and software.
While storage virtualization can bring better utilization of storage resources and the ability to migrate data between different tiers of storage, you need to look at those capabilities and decide whether or not you need to deploy storage virtualization to achieve them. If, for instance, you are only buying virtualization to determine the utilization of your storage arrays, there is other software than can do that.
Management of a single homogeneous environment may be easier with software supplied by the vendor for its storage platforms. Management of diverse storage platforms such as those from IBM, EMC and Hitachi may be simplified by storage virtualization since all data from the different devices would be aggregated into a single pool that can be managed from one management console.
Storage virtualization comes in three forms: host-based, network- or fabric-based, and array-based.
Host-based virtualization has been around for a number of years and is relatively inexpensive compared to network, fabric-based and array-based virtualization. Host-based virtualization is characterized by Symantec's Veritas Storage Foundation or Brocade's Tapestry StorageX. It is, however, often plagued by lack of scalability. As a storage virtualization environment grows, more servers are needed to host virtualization – each server requires its own operating system and host virtualization license, maintenance and software overhead.
In-network or fabric-based virtualization is getting a lot of interest lately because it allows the attachment of any variety of host computer and almost any vendor's storage array. Network or fabric-based virtualization is represented by IBM's SAN Volume Controller and EMC's Invista. In network or fabric-based virtualization you need to decide whether to adopt an in-band or out-of-network configuration.
Controller or array-based virtualization is characterized by Hitachi Data Systems TagmaStore Universal Storage Platform. Controller-based virtualization does not require you to insert another appliance into the mix, but does assume extra work for the controller.
Storage virtualization requires both additional hardware and software be added to the network. Host-based virtualization often requires software drivers on all host computers that attach to various storage devices. Fabric-based or in-network virtualization requires an appliance which attaches to the Fibre Channel switches. Array-based virtualization may require the addition of a separate storage array to virtualize the storage resources.
You need to also look at any changes to your network that virtualization require. For example, that may come in the form another brand of Fibre Channel switch or a pair of servers you ordinarily wouldn't need to have.
You need to decide whether you are going to be virtualizing storage devices from one vendor or those from two or more storage vendors. Some forms of virtualization are limited to only those storage devices of their own manufacture. Others, like Hitachi's TagmaStor USP can virtualize data from a variety of vendor's storage arrays.
If, for instance, you want to migrate data between different storage platforms from a single vendor, you may not need storage virtualization at all, as migration software from that vendor might be a cheaper end to the same goal. If you want to migrate data from several platforms to a single one or migrate data stored on storage arrays from diverse vendors to a single system, then pooling that data with storage virtualization may make migration easier and a much less complex process.
In choosing the type of virtualization to deploy, you need to watch out for vendor lock-in. Some array-based virtualization such as that from Hitachi Data Systems, requires you to add a Hitachi storage array to the network in order to virtualize any data. You also need to be sure that the functions you need to perform -- such as initial data migration -- do not require you to switch to the vendor's own migration product. Locking yourself into a vendor's replication, migration or management capabilities can be a costly affair.