The Best Is Yet to Come: Radical Changes in Information Storage and Processing
Storage area networks are being deployed in large numbers today and are being used to solve real business problems. But, compared to their ultimate potential, the surface is just being scratched. This section touches on some of the more radical changes in distributed information storage and processing enabled by SANs.
Best-in-Class Computer Systems
SANs enable the interconnection of many servers to many storage devices of different types, including disk arrays and robotically controlled tape drives. Who is it that makes these devices and how is it that they all work correctly together?
Both of the market-leading SAN interconnect technologies, Fibre Channel and Ethernet, as well as the emerging Infiniband, are standards-based—that is, their specifications are defined by voluntary associations of (sometimes competing) companies and are easily available to the general public, rather than being the private property of one company.
Since the same interconnect specifications are available to anyone, in principle, any company should be able to build devices that interoperate with any other company's complementary devices, creating heterogeneous storage area networks. This, in fact, is a stated goal of the storage networking vendor community.
This goal has only partly been realized today. No vendor has developed all the components required to build a complete SAN, but most vendors are engaged in partnerships to qualify and offer complete SANs consisting of the partners' products. Today, consumers are well advised to make SAN purchases from a single vendor or, at a minimum, to choose a primary vendor for major components such as RAID subsystems and accept that vendor's recommendations for other SAN components. The industry is working toward a mode of operation similar to that of the LAN industry, where component interactions are sufficiently well understood that users feel relatively comfortable making purchases at the component level. Much energy is being devoted to interoperability, both in terms of making standards more precise and in terms of validating components against standard test suites rather than against individual complementary implementations.
Smarter Storage and Appliances
While the cost and performance of online storage have improved pretty dramatically over the last decade, its functionality has remained pretty much the same. While other major computer system subassemblies, such as network interface cards and video controllers, have become more intelligent and increasingly autonomous, a disk has pretty much remained a disk and a tape has remained a tape for the last decade.
It's only partly coincidental that with SANs becoming popular, storage device functionality is changing as well. Today, intelligent storage devices that can transfer data among themselves without passing it through a setter are starting to be delivered. Peer-to-peer I/O improves quality of service by halving the bandwidth required to do bulk data transfers such as backup or replication of large file systems and databases. This will ultimately enable application designs that rely more on bulk data transfers, for example, to create more frequent backups or data replicas.
With intelligent storage devices, other advanced functions are also possible. For example, server-based hierarchical storage management software that compresses infrequently used data is available today. Implementing this capability in an enterprise RAID subsystem is technically feasible.
For other examples of how storage device intelligence might be exploited in the future, one need look no further than the storage middleware segment. Backup within the storage subsystem, clusters of storage subsystems, and autonomous global data replication are all possible. Some vendors have introduced these capabilities in storage appliances that are nothing more than computers dedicated to managing large amounts of storage and I/O capacity. With these storage appliances, the distinction between a computer and an enterprise storage subsystem essentially comes down to whether or not the box runs applications.
Heterogeneous Computer Systems
Today, it is quite common for users to partition the storage in large RAID subsystems mid allocate the partitions to different computers made by different vendors and running different operating systems. The computers share the RAID subsystems' common resources (power, cooling, processors, internal bandwidth, cache, etc.), but each has its own private online storage capacity.
With a SAN, however, it should also be possible to share storage and data across computers. In principle, computers from Sun, Hewlett-Packard, IBM, and others should all be able to access data stored in one big enterprise storage subsystem made by EMC, Hitachi Data Systems, HP, or any other enterprise storage subsystem vendor.
Today, data sharing among heterogeneous computers is possible with network attached storage or NAS devices. With other network storage devices, most notably enterprise RAID subsystems, it is not generally possible to share data among computers of different types (although some clustering technologies support data sharing among computers of the same type).
Many difficult technical problems must be solved before different types of computers can share data. By far the greatest of these is that each operating system and file system uses its own unique disk format, simply because there has never been any motivation to do otherwise. SANs provide this motivation because they interconnect the storage of different types of computers running different operating systems. Today, therefore, both academic and industrial researchers and developers are busily working toward the goal of universally accessible online data files.
Data Storage as a Service
Online information storage has traditionally been something that enterprises have implemented and managed for themselves. Whether so inclined or not, information services professionals have been forced to develop storage expertise, as it is they who are responsible for safeguarding the organization's information assets arid keeping them accessible.
SANs are enabling a new generation of storage service providers, companies whose business consists of storing and managing the data from other enterprises. These companies take on the hard problems of providing secure data access at contractually defined quality-of-service levels. In the not-too-distant future, many enterprises may literally get their data storage from a plug in the wail, much as they get their voice and data communications services from external providers today.
Similarly, the long-distance connectivity of SANs will someday enable the secure interconnection of storage networks belonging to enterprises that do business together or that share a specific market. Storage intranets will allow partnering enterprises to replicate or even share information without risk of its being divulged to (possibly hostile) outside parties.
Widespread adoption of these new modes of operation would clearly require significant technological development security, as well as changes in user attitudes. They may or may not become prevalent. The point to be made here, however, is that SANs are enabling developers, entrepreneurs, and users to break out of the box as they design enterprise information storage and processing infrastructures. There are all kinds of possibilities. We probably haven't even envisioned the ones that will be the success stories of five years from now.
