Keep This in Mind

A new way to manage storage problems is available.

by / July 31, 1999
W ith storage-performance and storage-space dilemmas all too common for sites running memory-killer applications like imaging, transaction processing, multimedia, data warehousing and Internet downloads, storage area networks (SANs) are widely discussed and equally in demand.

As networks continue to expand and applications churn more and more data, increasing amounts of high-performance storage subsystems are needed to keep pace. SAN solutions address these rapidly growing storage needs.

As processor and application performance have increased exponentially over the last decade, UNIX and Windows NT distributed platforms have largely replaced data-center-centric computing models. But during this same time span, improvements in file access and data-transfer techniques have been minimal.

Now, they have reached the critical point where server-to-storage communications have become the major roadblock to improved systems' availability and performance. SAN technology is changing this, deploying mainframe-class, host-to-storage functionality to the enterprise network environment at open systems price-points.

A SAN is a dedicated network-type communications link between multiple servers and storage devices. Its high bandwidth and high availability provide improved server-to-storage efficiency and throughput while freeing the LAN from the large packet data transfer chores and the bandwidth-intensive overhead associated with storage tasks. The removing of storage-access functions provides a big boost to overall network performance.

Now ...

In the traditional approach, a storage device (e.g., disk, tape, RAID) is typically attached to an individual server via a point-to-point connection, such as a SCSI bus. With this one-to-one structure, each server utilizes its own unique data-management architecture, segregating storage-management tasks and making centralized control nearly impossible in an open-systems environment. In fact, when a site requires additional storage capacity, MIS administrators often turn to offline expansion on each server as the safest and simplest solution.

Additionally, file accesses and data transfers between storage devices must travel the busy LAN pipeline, straining bandwidth resources and resulting in sluggish response times across the network.

Overall LAN degradation caused by extensive server-to-storage communications over the network is common and not easily remedied in present SCSI-bus configurations. Due to restrictions in cable lengths, bandwidth and connectivity, MIS administrators are hard pressed to break this data-traffic logjam.

... and Later

This is the year SANs cross the chasm between early adopters and general market acceptance. Two years ago, less than 10 percent of IT administrators in a distributed LAN environment understood storage area networks. Today, SAN stories fill the cover pages and storage sections of most IT publications and SAN benefits are being discussed at all major industry events.

As SANs are integrated into existing enterprise networks, users gain faster access to gigabytes of data, a higher level of data availability and a more reliable, fault-resilient server-to-storage-device hardware path than was previously possible.

Most early SAN deployments concentrated on enhancing storage-to-network data transfers by replacing SCSI buses with Fibre Channel I/O paths for communications with disk storage devices. These installations received the immediate benefits of increased bandwidth (100MB) for faster file throughput over longer distances (10 kilometers) than was possible with SCSI-buses (80MB and 25 meters, respectively).

Today, new software is enabling additional SAN benefits like remote tape vaulting and disk mirroring, backup/restores with dynamic tape drive allocation and high-availability clustering. In the future, SANs will turn storage into an enterprise-wide commodity allocated to applications when and where needed.

With a SAN, storage devices are moved off the server backplane and onto their own high-speed network. When this is done, overall LAN performance improves because it is freed from bandwidth-intensive storage functions. Any host on the network can access any storage device and its stored files without interfering with LAN traffic.

Various storage units (disks, tape libraries, RAID) can be mixed and matched and they can all communicate with multiple hosts and with each other. The storage subsystem is isolated from the operational limitations of the server and LAN traffic constraints. Alternative data paths are established between storage devices and servers, fostering fault-tolerant operations.

Storage components can be easily managed and centrally administrated and storage is externalized, meaning storage capacity can be scaled in accordance with storage needs and is not tied to server performance.

The result is the SAN creating a shared storage repository attached to a high-speed pathway, a channel optimized to handle the fast throughput and large-file-size transfer requirements demanded by today's high-end storage tasks associated with applications like video, CAD, OLTP and data warehousing.

The SAN will eventually, when the software becomes available, allow for universal data sharing across all platforms on the enterprise-wide network and handle high-bandwidth storage management functions like archive/retrieval, backup/restore and disaster recovery.

What Makes a SAN?

The SAN consists of servers, network interface cards, Fibre Channel interconnects, storage devices and either copper or optical cabling. Intelligent switches make up what is commonly referred to as the "fabric." Most SAN implementations are expected to employ Fibre Channel fabrics, as they offer the bandwidth, scalability, distance, reliability and investment protection sought by many users.

SANs are not necessarily synonymous with Fibre Channel. But so intriguing is the potential performance promised by Fibre Channel vs. other bus-connections (such as SCSI and ESCON) that Fibre Channel has vaulted to the forefront of SAN developments. Switched Fibre Channel fabrics provide unparalleled levels of flexibility, high throughput and inherently fault-resilient access paths, rivaling mainframe-level host-to-storage communications.

These switched fabrics support modular scaling of devices and up to 16 million device addresses. The channel handles multiple protocols and 100 MB/second, full-duplex bandwidth over each switch path. The Fibre Channel switches can be cascaded to provide dramatic capacity and aggregate performance scaling.

Fibre Channel's purpose is to address server-to-storage interface limitations. Its throughput capability greatly enhances information transfers at any site where high-end applications (e.g., imaging, video, OLTP, databases and CAD/CAM) are outpacing storage-system performance. The extended host-to-storage device connectivity distances possible with Fibre Channel, over six miles, is ideal for applications like remote backup, archiving and mirroring for disaster- recovery purposes.

In addition, the Fibre Channel support for long-distance ATM connections is enabling a whole new class of enterprise-wide storage applications.

Fibre Channel switches provide for highly resilient modes of operation, storage-device scalability, and easy hot-plugging/unplugging of devices. This means that reconfiguration of the storage component is simple and does not affect SAN or LAN operations.

Security and isolation between functions is another advantage of switched-fabric SANs. Fabric switches provide the self-learning and self-healing features to tackle any array of storage-intensive management tasks like full-network backup, remote vaulting and hierarchical storage. Fabrics composed of multiple switches enable Fibre Channel networks to grow to very large sizes, enabling continuous operations with centralized control over distributed storage resources.

SANs Deliver The Goods

As business operations keep expanding, more mission-critical data must be kept online and available. In fact, the amount of data to be stored and managed is now doubling almost every 12 months. This data eventually requires additional server-storage space. The most common non-SAN alternatives to meeting these increased storage needs are adding another individual server disk, replacing the existing server disk with a larger unit or adding another server "box."

These are all expensive and time-consuming options. With a SAN solution, storage capacity can be scaled as needs dictate without interrupting network service, without reconfiguring hassles and without regard to the server's performance or enclosure limitations.

Administrators want a storage solution they can use with existing networks, thus protecting investment in current IT systems. Everybody wants a storage system that is easy to manage, scale and integrate into their existing structure. SANs do all of this. They offer centralized storage management, independent storage subsystem scalability and they are quickly integrated into most major network operating platforms such as UNIX and NT.

By enhancing storage area networks with the appropriate software, the SAN addresses the need for reliable, multiple server-to-storage access and fast LAN-storage device file transfers. The SAN also makes an ideal off-LAN I/O channel to handle the tasks of online and remote backups across the enterprise network.

Fibre Channel Switched Fabric SANs guarantee high data availability at maximum bandwidth across many fault-resilient configurations. By eliminating single points of failure along the channel, near 100-percent uptime for file access is achieved.


Ron Levine is a writer based in Carpinteria Calif. E-mail