Virtualization technologies are most widespread in the server world where they are harnessed to bring about greater efficiency and cost savings.
In this exclusive interview, Business Technology Asia speaks to Hu Yoshida, Chief Technology Officer, Hitachi Data Systems to find out more about the lesser-known benefits of storage virtualization. Hu also provides some tips on factors that companies should consider when shopping for a storage virtualization solution.
1) Most people are more familiar with server virtualization. How important is it to virtualize storage and how will it enhance resource allocation?
With data storage growing at exponential rates, traditional storage networking, provisioning and application consumption methods are costly and inefficient.
Storage virtualization is one way to consolidate across all your available storage resources, which enables businesses to optimize their storage allocation through technologies like thin provisioning and tiering. By reducing the wastage caused by over-allocation, this increases the available storage pool, allowing businesses to delay costly storage purchases.
Contrary to popular belief, the fastest growing storage costs are not the initial capital expenditure (purchase the storage infrastructure), rather they are the operational cost related to managing that infrastructure. These include real-estate, power, cooling and manpower management costs.
Storage virtualization is one of the key to preventing these storage-related operational costs from spiraling out of control.
2) What is dynamic tiered storage and how can it help companies reduce costs?
Tiered storage is a strategy to distribute different classes of data across differently valued storage elements. These tiers are distinguished by differences in cost, performance, reliability and feature sets.
Although storing everything on the same class of storage may seem easier, it creates unnecessary costs. For example, email archives in most industries do not require the same performance level as business-critical applications. In this case, the data can be moved to tier 2 or tier 3 storage, lowering overall storage costs.
A tiered storage approach helps to move data based on volume groups and their application requirements, relieving companies from the arduous task of micro-managing the classification of all data.
3) Is storage virtualization more suited for large enterprises? Would it benefit medium and small-sized businesses that do not have tens of terabytes of data? Would they be able to afford it/justify the costs involved?
The benefits of storage virtualization can be felt from SMBs through to the largest enterprises. It is important to note that SMBs these days grapple with much the same issues of data growth and compliance regulations as large enterprises – typically with less available resources. While the sheer scale of enterprises result in larger absolute cost-savings, the OPEX savings that smaller organizations realise in the long run, typically outweigh the initial investment.
4) How are companies responding to storage virtualization? Are they keen on investing in this area?
While we are seeing more companies in Asia adopting storage virtualization, this rate is notably lower than their counterparts in the US and Europe. However, the economic benefits of storage virtualization are something most businesses cannot ignore given the current economic climate.
5) What are some of the factors inhibiting them? How can these be overcome?
The perceived cost of storage virtualization technology is one of the inhibitors to adoption in the region. One thing that CIOs and IT managers are starting to realize is that capital investment in infrastructure is but a fraction of the total costs of the technology over its lifespan. In many cases, the cost savings of virtualization makes up for implementation costs within the first three years.
Another common deterrent to a virtualized model of storage is the perceived security risks of multi-tenancy within the same storage resource. However, safe multi-tenancy within the same storage is possible through technologies like SAN zooming in the switch and LUN masking in the storage.
With IT budgets and headcount staying flat or decreasing, organizations need to find more efficient ways of managing their information assets. This is motivating more and more CIOs to re-look storage virtualization as a financially viable technology for the future.
6) Storage virtualization pools resources together. Do companies run the risk of revealing their confidential information to others?
Safe multi-tenancy, is a term that few people seem to understand. It refers to multiple users sharing the same resources safely.
For example, a hotel provides beds for many users or tenants. Safe multi-tenancy would be a hotel where each tenant had their own room with a lock on the door, versus a youth hostel where tenants share a common room and safety is dependent on the good behavior of the tenants.
In the same way, safe multi-tenancy is required for storage users who share the same storage resources, like a SAN. Unlike DAS or Direct Attach Storage, where a user’s storage is directly attached to the user’s host server and is separated from other servers by an air gap, SAN’s create a network where any server can get to any storage that is connected to the same SAN. Safe multi-tenancy is provided by SAN zoning in the switch and LUN masking in the storage systems. This type of protection is limited and makes LUN address assignments difficult since one user may get access to LUN 0, 1, 3; and the next user would get LUN 4, 5, but not LUN 0 which is needed as a boot address.
Safe multi-tenancy becomes even more important when virtualization is layered on top of the SAN. While switch zoning still works, LUN masking is difficult since that is done in the storage controller and not in the SAN based virtualization appliance. Hitachi avoids this problem by implementing storage virtualization in the storage controller where Host Storage domains provide the protection for safe multi-tenancy.
While safe multi-tenancy can protect users from accessing each other’ data, it is not enough. There is still the danger that one user may dominate the use of shared resources and impact the performance of other users. For instance one user may kick off a data base reorganization which will suck up the shared cache, rippling through rows and tables, just as another user is trying to service a high performance transaction.
In order for users to get the right Quality of Service, QoS, when they need it, there must be a way to logically partition the use of common resources like cache and be able to change that partition dynamically based on policies that are triggered by time or events.
Logical partitioning goes hand in hand with safe multi-tenancy and is important when many users are sharing the same physical storage resources. This is true for a SAN, but is especially true when this sharing is done through a virtual abstraction. If you are considering storage virtualization, ask the vendor how they provide safe multi-tenancy and logical partitioning before you commit.
7) What are some of the key things that companies need to consider when shopping for a storage virtualization solution?
It is important to understand that storage virtualization is a process. Starting small – in one part of your storage environment – will yield some benefits, which will compound steadily as you extend the virtualization to other applications.
The planning process is crucial in a storage virtualization implementation. Each organization’s IT requirements and priorities are different. You need to establish how critical an application is to the business, as well as its availability and performance. Once the objectives and priorities are understood, the storage architecture, storage management standards, policies and operations procedures can be determined.
The next step is to determine the method of virtualization that is right for the nature of the business (i.e., virtualization in the server, in the storage network or in the storage controller). There are different issues associated - primarily around data migration and downtime - which need to be explored fully prior to selecting the virtualization technology.
