IT managers typically have two queries on the storage front. The first is about future-proof storage architectures, whereas the second is about how to achieve more with less. Storage virtualization is a part of the answer to these queries. However, storage virtualization is making a slow but steady progress in India, with only large enterprises evaluating the technology at present.
Storage virtualization is basically about the pooling of various storage devices on to a central console. The approaches for deploying storage virtualization (in-band, out-of-band and split-path storage virtualization) are determined by various factors. In-band and out-of-band storage virtualization methods are more popular.
There are several perceived disadvantages with in-band and out-of-band storage virtualization approaches — the debate continues. Most vendors who use the open system approach for storage virtualization use types of block level aggregation. Out-of-band storage virtualization is suitable for file-level aggregation in large high performance computing (HPC) requirements. The file concentration is very high in HPC environments, which essentially have Ethernet- and file-based operations. Out-of-band storage virtualization implementations are typically custom-built.
For in-band storage virtualization, the appliance is placed in the data path. In-line storage virtualization typically uses block level aggregation with the following approaches.
a) Host-based: This approach uses an intelligent host, which has the required bandwidth for aggregation. Storage virtualization is integrated into the host.
b) Storage network-based: This is the most popular approach for in-band storage virtualization.
c) Device-based storage virtualization
In-band storage virtualization can be started at a very small level, though implementations are very different and custom-built. Keep in mind the following pre-requisites if you plan to opt for in-band storage virtualization.
1. Assess your storage architecture's design and implementation to determine the required multi-pathing. Depending on the implementation, multi-pathing may become a charged item or a free item, thus leading to a hidden cost.
2. Look at your SAN design and decide how many SAN fabrics are involved. For example, does the project involve 1, 2 or 4 fabrics?
3. Does your virtualization engine provide support for multiple SAN vendors? Does it talk to multiple SAN products available in the market? Also, does the storage virtualization engine support all the available standards and switches?
4. From a storage virtualization solution perspective, you have to look for features such as whether replication can be performed from the virtualization engine. If the replication is still based on arrays in the virtualization engine, then it will be a very complicated and extensive implementation. So if storage virtualization controllers can perform replication on the other side, then it saves complications and costs. Does the virtualization engine support data movement? Another important thing to look is the performance numbers.
5. It's essential to ensure that the personnel handling your storage virtualization setup has in place his basic storage knowledge. A basic training session about the tool is also imperative to get the best out of your storage virtualization deployment.
About the author: Sandeep Dutta is the vice president of IBM India/SA's storage, systems and technology group.
(As told to Jasmine Desai.)