This article can also be found in the Premium Editorial Download "Storage magazine: Five ready-for-prime-time storage technologies."
Download it now to read this article plus other related content.
Today's ILM product initiatives generally amount to little more than repackaging old products with new names. But with so many companies investing in ILM technologies—a concept that's fundamentally sound and necessary—it's a foregone conclusion that all data will be managed with ILM in the future. When that will happen is a matter of debate, but the essential concepts are on the way and some basic products are here.
It will be extremely difficult to meet all three ILM requirements for the vast pools of unstructured data that make up most deployed storage today. But all three are available for some applications. E-mail systems, document repositories and databases, for example, contain the essential meta data to feed an ILM system. And new "extender" products, from EMC/Legato, KVS and others, enable the essential mobility element for these systems. If you have an automated e-mail archiving and retrieval system, you have ILM for at least one application.
HSM software can be considered simple-minded ILM for file systems. It typically monitors the only metric available for files—age—and realigns data protection by selectively moving files from one type of storage to another. But HSM has shortcomings in all three areas. First, file creation or the modification date isn't really a valid metric for data value. Another problem is the file-level granularity of HSM; many applications store data
Perhaps the most intriguing alignment technology is content-addressable storage (CAS). By breaking the bonds of physical connectivity (think SAN) and file-system layout (think NAS), CAS would theoretically allow an entirely virtualized pool of objects, complete with the ad hoc metrics of value and data movement. While today's CAS systems are fairly basic, vendors are likely to catch on and begin offering CAS networks with tiers of storage, per-record data protection and automated movement on changes.
The biggest hurdle to CAS, however, is beyond the realm of hardware vendors. Applications must be ready to address data as CAS records rather than as files or blocks, and precious few have this ability. But again, there's a technology fix on the horizon. Most operating system vendors have been laboring to build new database-type file systems. But these keep getting pushed to the next generation or remain obscure—such as the Apple Newton's "soup," Windows Cairo OFS, BeOS BFS and Windows Longhorn WinFS. However, it's likely that a database-enabled file system will eventually allow an operating system to interact with a CAS network directly. This will be as big an enabler of ILM as any yet seen, as applications will quickly take advantage of the new technology.
This was first published in December 2005