"SAN filers" may offer an option for consolidating enterprise-wide file services, overcoming some of the drawbacks of traditional NAS.
By Jon Toor
Thanks in part to simplicity and reliability, the network-attached storage (NAS) market is growing faster than the market for general-purpose servers. By consolidating data on easily deployed platforms, IT managers have found a simple solution to storage "bloat."
However, as demand for file-level storage grows, the proliferation of NAS servers creates a new form of complexity. Silos of data storage, with user groups tightly linked to particular storage devices, lead to uneven resource utilization and backup headaches. Although backup can be improved with data mirroring to low-cost disk (an increasingly common approach), this practice requires additional filers and increases storage management headaches. For many IT professionals, growing requirements have dramatically increased the cost and complexity of the once-simple NAS solution.
To date, the industry's approach to this storage dilemma has been incremental at best, with solutions deeply rooted in the NAS paradigm.
But new approaches will accommodate growth, reduce complexity, provide greater data availability, and deliver higher levels of performance. With technology elements drawn from storage area networks (SANs) and NAS, these approaches will reduce costs by eliminating single-vendor lock-in and by providing IT managers with the flexibility to choose the disk arrays that best match application requirements.
NAS gateways, which began shipping relatively recently, are a partial solution to some of the problems. "SAN filers" may be the next evolutionary step.
To better understand these two approaches, it is helpful to look at how file serving has evolved.
Introduced almost 15 years ago, NAS servers have succeeded by offering a combination of advantages relative to general-purpose servers, including reliability, heterogeneous file sharing, remote manageability, ease of use and installation, and performance advantages in pure file-sharing applications.
But because the underlying hardware and software of NAS platforms is generally derived from general-purpose servers, they share some weaknesses:
SAN-based asynchronous mirroring replicates data to a second, low-cost array. When needed, the mirrored copy can be immediately accessed by any SAN filer. The copy also acts as a backup-to-tape source.
Difficulty adapting to changing requirements—Storage performance bandwidth is one parameter that must be adapted on a real-time basis. Performance fluctuations can create challenges in NAS environments. Because users are tied to specific filers, each filer must be sized to the peak demand for that group. If requirements exceed available bandwidth, the only alternative is to migrate users and data to other filers.
Inefficient backup—The simplest approach to backup in NAS environments, attaching a tape autoloader to each NAS server, becomes unwieldy as the number of servers or sites grows. Centralizing backup simplifies the process, but does not address two fundamental problems: a shrinking (or non-existent) backup window, and zero tolerance for downtime when a data restore is required.
To remedy the backup headache, NAS vendors now offer data mirroring. Replicating a primary NAS server to a mirror NAS server provides two important benefits:
- Tape backup can be performed from the mirror NAS, so the task is offloaded from the primary NAS server. This eliminates the backup window issue; and
- When a data restore is required, the information is immediately available on the mirror server. In most cases, this eliminates the need to perform time-consuming restorations from tape.
However, mirroring is costly due to the need for additional NAS servers, and it can increase complexity.
To address these limitations, "next-generation" NAS systems provide capacity and scalability extending to hundreds or even thousands of terabytes.
They also include clustering and fail-over capabilities for greater availability. And load balancing between multiple servers provides performance scalability and improves allocation of storage resources.
However, two elements that are still missing are an adaptive architecture and an open, multi-vendor approach.
Adaptive storage architecture—File services should be considered an element of the overall infrastructure, and end users should have the ability to distribute workloads and eliminate single points of failure.
Open, multi-vendor solutions—Because no single vendor can cost-effectively address all storage priorities—cost, performance, capacity, and availability—users should be able to select components from multiple vendors in order to meet their specific needs.
A solution providing these two elements would reduce management costs through increased flexibility and simplified single-pool-of-storage administration. And it would reduce capital expense by increasing resource utilization and by allowing lower-cost storage arrays to be used where appropriate.
SANs provide these advantages (flexibility, heterogeneity, and scalability) today. However, SANs have not been widely adopted for file-serving applications, one reason being cost. SAN components have been expensive, though costs are dropping due to increased adoption and growing competition.
The second reason is that few solutions enable the use of open storage to support file services.
However, these issues can be addressed by NAS gateways and SAN filers.
A NAS gateway is similar to a conventional NAS server: It provides heterogeneous file services, is remotely manageable, and is easy to use from a user, group, and volume management perspective.
A NAS gateway differs from a conventional NAS server in its use of Fibre Channel-attached storage that can be shared among gateways.
(For more information about NAS-SAN gateways, see "NAS and SAN converging, but debates remain," InfoStor, March 2003, p. 26.)
SAN filers will offer another alternative for consolidating enterprise-wide file services on open storage.
Like NAS gateways, SAN filers provide heterogeneous file sharing to users on a LAN and store data on Fibre Channel-attached disk arrays.
However, a SAN filer is not based on existing NAS designs, as are NAS gateways. SAN filers are designed specifically to provide file services on open storage and can capitalize on the scalability and flexibility of SAN-attached storage to reduce costs and improve productivity.
Additional features include support for multiple storage tiers, real-time load balancing, and petabyte scalability.
Multiple storage tiers—The tiers enable the deployment of storage arrays with performance and availability features that match application requirements. Mission-critical data can be directed to the highest-performance storage, while less-critical data such as backups is stored on lower-cost arrays.
Real-time load balancing—Since multiple filers can simultaneously access the attached storage, SAN filers can automatically respond to changing performance requirements.
Flexible capacity deployment—SAN filers can address the entire storage fabric, allowing capacity to be easily re-deployed among filers. Whether existing capacity is being re-allocated or new capacity is being added, all filers can immediately "see" and use the new storage.
Improved backup—SAN filers achieve quick backup and reliable restore with a multi-tiered mirroring approach. The primary array is mirrored to a backup array, which is then backed up to tape (see diagram on p. 34).
Mirroring is accomplished over Fibre Channel rather than over the LAN. This improves performance and eliminates the need for a second filer to access the backup array.
Users and applications can access the backup data via the primary filer, just as if no failure had occurred.
Jon Toor is director of marketing at ONStor (www.onstor.com) in Los Gatos, CA.