Distributed file systems enhance NAS

New file systems will improve scalability and storage management.

By Jeff Tabor

Forrester Research studies show that storage management can cost a company seven times the original equipment outlay—a fact that every IT manager understands all too well by now. This situation was caused in part by technology advances that have spawned a broad spectrum of storage alternatives, from commodity plug-and-play devices to complex storage fabrics.

Unfortunately, none of these solutions are easy to manage. Faced with increasing demands for more storage and faster access to it, storage administrators have amassed growing collections of servers, disk arrays, switches, and host bus adapters.

As new products with greater capacities and faster access get thrown into the mix, the result is an upward spiral of complexity, downtime, and management headaches.

Decreasing management costs

However, new storage technologies promise to significantly decrease out-of-control management costs. One approach is a distributed architecture that offers the simplicity of network-attached storage (NAS), the capacity and scalability of a storage area network (SAN), and the manageability that neither of these options currently allows.

On the surface, NAS seems like an ideal alternative to direct-attached storage: NAS servers are easy to set up, simple to maintain, and relatively inexpensive. You can support your application requirements by plugging in as many devices as you need, wherever you need them, on the existing network.

But the problem with current NAS architectures is scalability. Once an implementation grows beyond four or five devices, the NAS environment becomes difficult to manage. For example, you can't easily redistribute or consolidate data among multiple storage devices to improve efficiency, handle the fluctuations of large enterprise data pools, or support large database applications.

Click here to enlarge image


Next-generation NAS

Emerging storage architectures promise to shrink—and in some cases, eliminate—many of the elements that comprise the total cost of ownership (TCO) of storage.

Some of these architectures combine a distributed file system (DFS) with the traditional Ethernet-based NAS model.

A distributed file system can span hundreds of servers and allows IT managers to view and manage the entire environment as a single storage pool or logical name space. This architecture eliminates many of the performance and scalability drawbacks of traditional NAS, as well as the high costs and complexity typically associated with SANs.

The combination of NAS and a distributed file system (see figure) will help IT managers

  • Keep it simple—Retaining the simplicity of appliance-based technology, next-generation NAS devices can be quickly and easily plugged into existing network infrastructures. Also, by supporting a wide range of capacity and performance requirements, this approach reduces the soft costs associated with supporting a confusing mix of storage alternatives;
  • Manage many as one—With a distributed file system, devices are fully aware of each other. The file system automatically informs each unit of additions or deletions. The system tracks all file structures and subdirectories, giving clients transparent access to files anywhere in the NAS cluster. System/storage administrators can define a single global name space that spans multiple servers and can logically apportion storage to support business requirements;
  • Run enterprise-class applications—Next-generation NAS eliminates the capacity and performance constraints of traditional NAS. These architectures may deliver linear performance scalability, making it possible to run applications that call for large file sizes, large name spaces, and/or high throughput. This approach also simplifies data layout and other performance optimization tasks such as load balancing across the cluster;
  • Leverage industry-standard protocols—As does traditional NAS, these systems will support NFS and CIFS, as well as emerging block protocols such as iSCSI; and
  • Minimize downtime—The new architectures will virtually eliminate the user-access downtime generally associated with common management tasks such as data redistribution. Also, a clustered environment simplifies the setup of high-availability functions such as server fail-over and data mirroring.

Next-generation NAS architectures will support an application-centric approach to storage expansion, while delivering better expandability, performance, and manageability than traditional NAS approaches.

Jeff Tabor is a senior product manager at Spinnaker Networks (www.spinnakernet.com) in Pittsburgh, PA.

This article was originally published on November 01, 2002