Of late however, it has become rather more outgoing and ambitious, and decided to spread its wings.
So while storage virtualisation technologies have, historically, been limited mostly to the datacentre, and to mainframe and high-end storage array environments, their horizons have broadened, and now include solutions for the mid-market and even the desktop.
As a result however, they’ve also started having to sell their benefits in a rather more rarified and less forgiving atmosphere.
This in itself is a big ask.
While many of today’s storage offerings do promise some degree of virtualisation, in most cases, storage virtualisation itself isn’t yet a simple point solution that can be bought off the shelf. Just as with any other disruptive technology, the key lies in first defining terms and then understanding your requirement.
Storage virtualisation has many definitions and makes many promises. You could, for instance, describe it as dynamically allocating the most cost effective media for data storage in order to deliver the most appropriate access, back-up, restore, and archiving functionality for the organisation in question.
Put more simply however, it’s about sharing storage resources through a single view.
This means you can move data around different storage assets from different suppliers no matter where they are located physically, and allocate the storage to applications based purely on the storage policies and needs of the user.
This gives storage virtualisation the scope to deliver huge benefits. And so do current market dynamics.
The ongoing explosion in storage is conservatively projected by analyst IDC to hit compound annual growth rates of more than 50% over the next two years, and others put the figure at 100% or more.
Right now this growth is being driven predominantly by users who, faced with the need for more capacity, are simply throwing more disk drives at the problem. However, such tactics tend to be expensive; they overlook a number of data and storage management issues; and worse, they largely ignore the fact that existing storage assets are typically under-utilised by between 30 and 50%, according to most industry watchers.
How has this happened?
As is so often the case with IT, storage assets have had to grow in direct response to the requirement of the time. This has led to fragmented, under-utilised, inefficient ‘islands’ of storage dotted around the organisation.
The answer to simplifying all these over-complex storage environments that have built up in our businesses over the years? You guessed it. Virtualise them. And there’s no doubting the power of the argument. Virtualisation promises to drive down storage and management costs, to increase storage utilisation, to accelerate back-up and restore times, and to offer uninterruptible data access (i.e. business continuity).
It’s even being feted as the solution to everything from remote data management to reducing IT-related carbon footprints. However, as with any number of apparently simple solutions before it, the issues involved can, in fact, be pretty complex.
Moreover, when it comes to their data storage environments, most users won’t have the luxury of starting with a clean slate.
Most storage assets consist of different disk formats bought from different vendors with different controllers and different management software. At a datacentre level, there will likely be different storage arrays from different vendors too.
In addition, the networked infrastructure will probably encompass direct-attach internal IDE hard drives in old PCs and servers, and ATA and Serial ATA drives in newer systems.
There’ll also be different vendors’ external RAID boxes in different parts of the business, which will most likely house different disk formats, possibly connected over different network protocols. Network attached storage and newer disk formats and networks such as iSCSI and Fibre Channel may also be involved. And if the organisation is big enough, some whopping tape sub-systems too.
The type and the value of the data being stored also have to be considered. Some data will be static and of relatively low-value, so needn’t sit on expensive disks. Other data will be in constant use and demand and will therefore need to be instantly, continuously available. The storage system must therefore be able to make this and other important distinctions between varying data types.
Does that mean that virtualisation is only an option for big corporates with vast storage estates to manage, and with matching budgets? Not necessarily. At least not any more. These days almost every new storage product coming to market claims some form of virtualisation functionality. So now is the time to explore your options.
However, with alternatives ranging from storage array and network-hosted solutions through to software-, appliance-, RAID controller- and even server software-based products, it’s not always immediately clear where to start.
Beginning in virtualisation’s spiritual home, the datacentre, the vendor battleground is being contested essentially between the heavyweights Hitachi Data Systems (HDS), EMC, and IBM; although HP and Sun Microsystems also claim to play here.
In effect, each vendor’s solution offers an abstraction, with storage resources viewed logically and separate from their physical components and then managed accordingly.
However, there are big differences in approach, mostly around various “in band“ versus “out of band“ based offerings.
EMC’s Invista product is an example of an “out of band” solution, as is IBM’s SAN Volume Controller. i.e. both sit on the network and control data movement through network switches. HDS’s Tagmastore is “in band” – which means all the data passes directly through the device; usually a storage array.
There are arguments for and against each, most commonly relating to their relative performance, reliability, resilience, and ability to manage heterogeneous environments in multiple datacentres. The lines between these “in band” and “out of band” options are however, becoming increasingly blurred as “intelligent” directors and switches near the market.
Naturally, with more and more major virtualisation deals now coming up for grabs, each vendor’s favourite pastime involves finding flaws in their rivals’ offerings. In fact, they’re all quite vociferous in thrashing one another’s technologies.
Any more objective view however, ought to point out that HDS has won many plaudits for its Tagmastore products as they scale from department through to datacentre level and offer clear migration paths as they scale. It has also been in the market longest.
Then there’s EMC. At the high end there’s its Invista solution. Among its claims are openness, as it can sit on market-leading Brocade, Cisco, and McData switches. EMC Rainfinity is the company’s offering for mid-range to low-end installations. But it has, as yet, no clear scalability path from the low to the high end.
Elsewhere, IBM has its SAN Volume controller, another switch-based product for the enterprise. Its software controls the movement of data through a SAN and offers all the usual virtualisation benefits.
Moving further into the mid-market, appliance-based products tend to grab most of the attention. HP’s virtualisation approach is based around partnering with switch OEMs to deliver virtualisation on its StorageWorks storage appliances. Network Appliance is also active in the storage virtualisation appliance.
For its part, Sun Microsystems has a long term partnership with HDS and also offers mainframe level tape virtualisation through its Storagetek range. For the mid-range its approach is based on partnerships with other OEMs. Tape suppliers including Overland also offer virtualised tape solutions.
At the server software level there are companies such as Datacore, which sells server based software that controls storage assets and manages resources. In short it will control any disk type (SATA, ATA, Fibre channel) across most storage subsystems (EMC, IBM, Sun and more) by sitting on the server and dynamically allocating storage assets to applications.
So what’s right for your business?
That depends on your readiness for storage virtualisation, which in turn will depend on the size of your organisation and the scale of its storage infrastructure. Suffice to say though, that any firm with a sizable, heterogeneous or distributed storage infrastructure should at the very least be examining its storage virtualisation options.
Virtualising even a small part of your storage infrastructure is likely to deliver benefits and fast ROI. More importantly, it will also open the door to the next generation of data management.
Some Common Virtualisation Terminology
In Band
Virtualisation in which the device sits in the path of the moving data.
Out of Band
Virtualisation that’s usually delivered through an agent on a network switch. Claims not to slow the movement of data.
Tiered Storage
Seen as a function of virtualisation. Moving data to the most cost effective storage medium. E.g. storing vital data first on expensive, fast disk platforms, moving it to slower disks and eventually archiving it to cheaper tape.
Fabric-based virtualisation
A.k.a. switch-based virtualisation built on the argument that by putting the virtualisation layer on the network, different operating systems and data can be managed regardless of where they originate or their final destination.
Array-based virtualisation
Manages multiple arrays. Focuses on disk utilisation and separating the application from the platform where the data sits.
Software-based virtualisation
Sits between application and direct-attached storage device (or the network connection to the storage device). Makes virtual storage resources available to
the application.
