The lure of the SAN
By Amy Larsen DeCarlo
Network World Storage in the Enterprise Newsletter, 03/13/00
E-business is changing how businesses value information. Information has become a strategic asset that gives companies an edge over their market rivals. Companies use intelligence to identify new markets and make contact with prospective customers. In this media-saturated era, information itself is packaged and sold as a product. This makes the ability to supply users with fast access to stored information on a continuous basis absolutely crucial.
Companies are clearly coming to a crossroads in their storage implementations. With estimates for Internet storage capacity needs doubling every three months, IT professionals are hungry for a scalable solution to help them consolidate control of stored information. They often look to storage-area networks (SAN) as a better option to manage their information storage systems than distributed models.
Today, most organizations rely on a distributed storage model that uses file servers to process I/O requests from end users and other application servers. In this model, all requests for data go through the file server that owns the attached storage disks, and only one file server can tap data on a particular disk via a SCSI bus.
This model has several shortcomings. First, the amount of data a server can access is restricted to the number of disks supported by the bus, which limits the capacity of a single file server. Second, because the server processes each I/O request, it risks becoming a bottleneck. Third, this server model carries some daunting availability limitations, because only one file server is allowed to access a set of disks. If that file server or any of its SCSI connections fails, then users and other application servers lose access to the stored files.
This model carries other major disadvantages. Distributed file servers rely on the data transport network to run backup and recovery operations which can eat up bandwidth and slow normal network transmissions to a crawl. Finally, this decentralized setup is difficult to manage from both a logical and a physical perspective. File server based storage systems are distributed throughout the enterprise, so it is often difficult to assess current and future capacity needs. And because these servers use a parallel cabling scheme to link the file server to the disk array, they can also be cumbersome to set up and manage.
SANs promise to mitigate the problems that plague conventional file servers, largely through consolidation of control. These specialized storage networks claim higher availability, faster performance, centralized management, and by their architecture, the capability to remove bandwidth-intensive data backup and recovery operations from the LAN. This frees up the LAN for normal data communications and ensures smoother back-up operations.
Using high-speed transports like Fibre Channel, SANs offer a high-performance network optimized for moving storage data. SANs also make way for new storage implementations like LAN-free backup. And, because Fibre Channel can support distances of up to 10 kilometers, SAN devices can be widely distributed, but also centrally managed as one network.
Yet, as was the case with LANs in their younger years, SANs are still developing. Vendors are still working out major product interoperability issues, while companies deploying SAN technology struggle with how to merge the very different worlds of storage and networks and manage both together. Ultimately the hope is that, like LANs, SANs will develop into a mature and highly manageable solution that supplies substantial benefits at lower costs.
Given that storage deployment and ongoing support costs can total 10 times the acquisition price for the equipment, the consolidated management capabilities of a SAN may deliver the biggest benefit to business.
nwfusion.com |
