Information economy will see massive changes in how companies will view, manage, and store their data. This has been an unanimous observation by each and every player in the storage arena. Without any prejudice, everyone affirms the fact that today, CIOs are aware of the need for flexible and quick access of data, anytime, every time, but the problem faced is how to accomplish that. It has become even more difficult in a scenario where most servers have their own storage capacity. The issue is about managing the system itself—the system of spread storage, space, scalability, redundancy, reliability, backup, availability, security, performance, inter-operability, maintenance, investment protection, and return on investment.

The issue further compounds with the emergence of newer architectures and new equations among the vendors. Especially, when on one hand, people talk about the need to separate server and storage functionalities, and on the other hand, new classes of device making to the market place with combined functionalities. These developments have perplexed even the analysts, then one can imagine what it would mean to the customer. The questions that arise are: Can products that resemble servers and give fast access to storage, be classified as servers or as storage? Or do these fall in the category of multi-utility? What if these can also handle caching and load-balancing properties? The mute question therefore is should a server, which should be freed of processing I/O requests, be combined with storage, or should the server handle the tasks like computation and brute-force number crunching and let the storage run separately?

Reinventing Architectures

Since IT architectures weren’t designed for the new ways in which businesses use data, especially when enterprises want to embrace businesses in close proximity with their customers by integrating applications and data, want to create ‘no-wait’ kind of a customer experience, leverage the Net, etc, it is becoming imperative to look at new architectures. Further, the pressure on IT managers to control costs has surmounted infinitely. Let us look back at what had happened recently. The purchases of storage hardware and servers were done in unison, as the storage devices were mostly attached to server. In such a model, resources remain unused as the device could not be shared. Further, the cost of management of server-attached storage is not only complex, but would also mean an increase in costs for managing them. Similarly, with increasing data, the need for bandwidth-intensive backup and recovery in quick time, and 100 percent availability, too, have become critical. It is in this scenario that storage networking and centralizing of storage—different types of storage devices and subsystems on one or more networks—are assuming significant proportions. The advantages cited are that more people can get access to more information and economies of scale, as the big storage device is understood to cost much less than several smaller ones, and managing a centralized system is much more effective than a dispersed one.

Most of the vendors point to one aspect—management costs get lowered with the newer architectures. Alvin Ow, regional SE manager, Veritas Software (Singapore) Pte Ltd, alleys the fear that storage is a recurring cost. In fact, new architectures would mean labor and cost savings. He points out to a Forrester Research Inc.’s study according to which as against a traditional architecture, the new SAN would bring about a 27 percent drop on cost and labor (see graph). Further, the centralized storage management is understood to increase the number of GB managed per storage manager by about seven times as against a decentralized system. And at the same time, the management costs that account for more than half of the total storage budget currently, in a decentralized system, would drop to one-sixth of the total storage budget.

Further evolutions on optical connection front, IP SANs, and storage services are also expected to lower the cost factors. Most of the people we spoke to admit that the future of storage networks is optical. They cite that optical networking is a quick and an easy solution to the distance and capacity limitations of today’s SANs. They say SCSI method of linking computers to storage extends to about 25 meters. But fiber and mainframe channels (Escon and Ficon) have a reach of more than 10 km. And with DWDM, a fiber channel link can reach to a minimum distance of about 100 km, without regeneration. Further, IP SAN could also be ?? because Fiber Channel over IP could address distance limitations and could be cheaper for those who lack in fiber channel know-how. And outsourced storage networks could be in a vogue, especially with DWDM getting wedded to storage. The service providers will offer off-premises backup to ISPs, ASPs, and others, who require large amounts of storage.

The name of the game for storage will be "carrier class solution"—forever-on, adaptive, dynamic, secure, seamless, efficient, and cost-effective.

Ch. Srinivas Rao