Why data storage is like inventory management

Network administrators wrestling with data storage management can learn a thing or two from a different kind of storage management: the management of physical product inventories.

Like inventory, data storage can be centralized or decentralized. When storage is centralized, management is generally easier and costs less. Part of this is just basic arithmetic; if I have fewer devices to manage, I need fewer licenses and tools to manage it. It is also easier for a storage manager to handle more storage when it is centralized than when it is decentralized, which reduces costs of labor associated with storage management as measured per megabyte. Think of it as walking through only one warehouse to check the inventory as compared with traveling between multiple warehouses. Storage-area networks (SAN) can make multiple storage systems appear as one pool of storage resources, representing a warehouse of capacity.

A drawback to centralized inventory is the need to transport goods farther when they are needed. Data, like inventory, should be located close to the client, unless network capacity is available. As streaming video and graphic files are added, this becomes an important consideration. The critical point to remember here is that the cost to transport data, like the cost to transport inventories, should be considered before making any changes to storage locations. Very few warehouse managers have been promoted when they closed warehouses and drove transportation costs through the roof or, worse yet, choked the delivery system to the point of failure! SANs again can provide the necessary bandwidth to ensure that storage deliveries can be made over distances not available with traditional storage technology.

Decentralizing data is desirable in some cases to place data closer to where it is needed. One approach to this is to cache data, making the information available where needed but keeping the burden for management centralized. This can be an effective strategy for data not subject to constant change. With the cost per megabyte of storage now in the 1 cent range and continuing to drop, caching copies of data close to the point of use presents a cheap alternative to constantly shipping data over the transport network. Inventory managers realize this too and may place small stockpiles of frequently needed goods throughout their distribution network, sometimes even in a customer's location. Designers of application systems should consider data caching when creating systems and use it whenever possible to support the goals of the enterprise infrastructure.

Also, like inventory, data frequently can go stale when unused for long periods of time and must be removed from the storage warehouse. This process is cheaper if done in a central location, as fewer searches are needed and less effort is required to clean out this old trash. Sometimes, a company wants to keep the data, but space in the warehouse is at a premium, and off-site storage must be used. Archiving and hierarchical storage management systems support these activities.

Perhaps the least understood benefit of centralizing data is the ability to do more with less. As with inventory, in calculating the amount of excess storage capacity needed to support growth, the ability to centralize can dramatically reduce the total storage requirements.

Inventory managers refer to this as "safety stock," and one of the critical components of safety stock is the variability of the demand. For data, this equates to how easy it is to predict the growth of the data over time, which can be quite a challenge for high-growth areas like information warehouses, messaging systems or e-business applications. By combining storage into larger centralized pools, the "law of big numbers" comes into play, which essentially means that things tend to average out to a smoother fit when you combine more and more numbers. In other words, you can get by with less excess capacity if it is all in one big system. This is because the growth changes in all of the applications will average out to less overall change.

When planning for centralized data storage in your organization, think about what it would look like in a physical sense, as though it were inventories of your company's most valuable assets. In considering any changes, understand - as would an inventory manager - the impact on the warehouses, on the transportation systems and, most importantly, on the customer.

Toby Miller is a senior analyst with Enterprise Management Associates in Boulder, Colo., a leading analyst and market research firm focusing exclusively on all aspects of enterprise management. He focuses on storage management and the growing storage networking technologies. With over 20 years in the IT industry, Toby has experience in many areas of technology including product development, marketing, technical research, software implementation, and technical architectures. He can be reached at tmiller@enterprise
management.com

Face-off: What's the better approach to IP storage?
Network World, 06/12/00.

Compaq describes storage management strategy
IDG News Service, 04/24/00.

Legato primes storage mgmt. software
Network World, 03/13/00.

Archive of Network World on Storage in the Enterprise newsletters