|
 ONStor’s Jon Toor explains why he believes that 2006 may be the year that use of virtualization technologies really takes off.
While we’ve heard about storage virtualization for years, wide scale deployment has not yet materialized. The technology was supposed to deliver benefits such as reduced management costs and better storage utilization, but for all the promise adoption among users remains low, its perceived cost and complexity remain high, and, frankly, the marketing hype has left users confused over the supposed advantages. In 2006, however, this will start to change. Virtualization will emerge this year as a useful tool, though it may not appear in quite the same form we’re accustomed to. To understand this transition, it helps to dig into what virtualization really is.
Virtualization means to decouple a physical asset from its logical representation. In storage, this means the physical location of data is not necessarily the same as the data’s ‘address’ (as a server would see it). The virtualization layer decides where the data will actually go. By decoupling physical assets (usually disk storage) from the logical, administrators gain the flexibility to put data wherever they like without having to reconfigure servers to do it. This flexibility can deliver benefits such as increased storage utilization through data pooling on a single array. Or it can enable ‘tiered storage’ by allowing the non-disruptive migration of data from one array type to another.
Traditional virtualization solutions accomplish this decoupling by adding a layer (the virtualization layer) within the system stack. The layer may be in-band (data flows through it) or out-of-band (data does not flow through it). But in either case, it’s an extra layer of hardware and/or software to buy, manage, and keep running. So, while virtualization can indeed enhance management, that flexibility comes at the cost of added complexity. For many users, that cost - plus the perceived risk - can overwhelm the potential benefits.
Nonetheless, the benefits of virtualization are real, so the question is how to gain those plusses without the extra burden of more stuff to manage. The answer is integration. A trend that’s gathering inertia is the integration of virtualization capabilities within other products such as disk arrays and NAS devices, which enhances management power within layers that already exist.
These integrated virtualization capabilities are starting to appear at three levels in the system stack:
Block-level Virtualization: This type of virtualization sits in the midst of the SAN (in front of the storage arrays) and masks the physical location of data from the device accessing this data. Standalone solutions available from vendors including IBM (SAN Volume Controller) and Datacore offer the benefits you’d expect: increased storage utilization, simplified capacity provisioning, and flexible data protection. But new solutions from HDS (TagmaStor disk array), and ONStor (Bobcat NAS Gateway) integrate similar block-level virtualization capabilities to deliver the same benefits with less cost and complexity. In 2006, expect to see more integrated solutions and added capability (such as advanced continuous data protection features) to further leverage the inherent simplicity.
Server (or Filer) Virtualization: File services requirements are growing. Many factors are driving this: growing client populations, new file-based applications such as content distribution, and a trend towards the use of file servers and NAS as a storage consolidation platform behind web and application servers. Combined, they push file services requirements harder than ever. In the face of this demand, processing power within those file servers is increasingly becoming a bottleneck.
What is Block-level vs. File-level data?
One way to distinguish virtualization solutions is the level of data they work with. Some work with block-level data (the type of information you would find on a disk array), while others virtualize file-level data, the type that file servers move over the LAN. So what’s the difference between these two data types? Most applications employ files, with data arranged in a format that the application readily understands. But when data is written to the disk, it does not get written in that same format. To centralize management of disk resources, most work applications through a shared ‘file system’ which acts as a librarian, keeping the data organized both as it is written to and read from the disk (caveat: some applications, such as databases, do directly employ block level data). This file system can reside either on a server or on a NAS device. If it’s on the server, block level data is written to the disk array, usually via either an FC or iSCSI interface. If the file system is on the NAS, file level data is written to the NAS device, usually via Ethernet. The NAS device in turn writes block level data to disk. A “NAS Gateway” is a device that maintains the file system and communicates file-level data with clients on the LAN, while storing data on external FC-attached disk.
An advantage of file level data is the ease of sharing information. Because the file system resides within the NAS, any device can access and understand the data. When the file system resides in the server, only that server can retrieve data from the disk. |
The traditional file server architecture – with servers tied to specific storage devices - made it difficult to add bandwidth when requirements grew. But new server clustering solutions fix this by allowing multiple devices to deliver file data from a single pool. This lets IT managers to more easily accommodate variable loads and to react to changes in demand. One standalone server virtualization solution, VMWare, delivers the ability to migrate workloads among servers. While effective, it is not specifically designed to meet the needs of storage users. Newer solutions address storage needs by integrating this capability within NAS devices. ONStor integrates server virtualization within its NAS Gateways, and NetApp offers it as an option on their Filers. In 2006, server virtualization will see greater acceptance as vendors enhance ease of use and add new capabilities.
NAS Virtualization: Multiple NAS devices can create a management headache. If one device becomes overloaded, the remedy is to move users and data to another device, a tedious and painful process. NAS virtualization eliminates this by abstracting the physical NAS device so any shuffling of resources that occurs among the physical devices remains transparent to users. Standalone solutions from Acopia and NuView provided this capability for existing pools of NAS. Other vendors such as ONStor now integrate similar features within NAS devices, offering benefits such as load balancing and seamless performance scaling. In 2006 you can expect to see more NAS providers integrating virtualization features that let storage administrators react to change without forcing changes on the client community.
Storage virtualization is indeed alive and thriving, but the trend is towards integrated rather than standalone solutions. This makes perfect sense when you consider how virtualization is deployed throughout the system world. File systems (which are a form of virtualization) and virtual memory have long played valuable roles; the needs of users and applications don’t always mesh with the constraints of the physical world. These solutions make it simple to match immediate requirements with available resources. In 2006, users will have increasingly simple and cost-effective options to deploy powerful virtualization tools in their storage environments.
Jon Toor is vice president , marketing , ONStor.
•Date: 21st Feb 2006 • Region: World • Type: Article •Topic: IT continuity
Rate this article or make a comment - click here |
