The price differences between the classes of storage are becoming extreme. For example as of this year, you could pay more than $120 per gigabyte for high-speed, high-reliability storage systems, or you could pay less than $7 per gigabyte for a low-speed, less reliable solution using ATA disks.
Obviously you don't want to try to run something like a business-critical transactional database on low-performance, low-reliability storage, but not everything in your enterprise is likely to be either business-critical or high performance. For example application development usually doesn't require a high-end storage network. Neither does less-critical data that needs to be kept, but will seldom be accessed.
This leads logically to the idea of tiered disk storage -- using different kinds of disk systems for various classes of tasks and saving money in the process. In a sense this is an extension of the notion of hierarchical storage management, except instead of shifting data to lower cost storage as it ages, you start out with the data on different classes of disk systems.
Tiered storage is a strategy that works best in a large enterprise with complex storage needs. For one thing, you have to have enough data in different classes to justify the expense of acquiring and managing different classes of storage devices. For example, Earthlink, one of the leading on-line services, has announced it is going to a tiered storage architecture using products from EMC Corp.. In the Earthlink scheme, the high end of the mix will be EMC's DMX1000 networked storage systems handling applications such as billing, data warehousing and Oracle databases which require more than three terabytes of data. The second tier uses EMC's lower-performing Clariion CX600 networked storage systems to handle applications such as call center management and customer relationship management. It is also adding a Celerra NS600 NAS system to provide file sharing to employees. At the low end, Earthlink will use CX600 systems with ATA drives for jobs such as quality assurance and backups.
However tiered storage can be cost-effective in even fairly small enterprises if there is enough spread between the performance needs of the various kinds of business data. For example, a business that has a small volume of data needing high-performance, high-reliability storage and a larger volume data that doesn't need either the performance or the high reliability can profitably use a tiered approach even if it is fairly small. A surprising number of businesses, from attorneys' offices to manufacturing plants, can fall into this category.
The key to a successful implementation of tiered disk storage is a thorough knowledge of your storage needs. You have to consider the kinds of data the enterprise handles, the amount of each kind of data and weigh cost savings of lower cost storage against the cost of installing and maintaining different storage systems. One very important consideration is management. Whatever kinds of storage you use should be centrally manageable, preferably with the storage-management software you already have. At the very least you need to be able to manage all of it with the same software.
EMC's announcement of the Earthlink deal is at: www.itweb.co.za/office/emc/0308080744.htm
Some additional information is available in the Earthlink customer profile included in: www.emc.com/customer_profile/customer_focus/cf_focus_summer2003.pdf
Stanford University is also considering tiered storage. A PowerPoint presentation giving its rationale is available at:
tss.stanford.edu/files/TSSAllStaff-June%202003-1.ppt.
Rick Cook has been writing about mass storage since the days when the term meant an 80K floppy disk. The computers he learned on used ferrite cores and magnetic drums. For the last twenty years he has been a freelance writer specializing in storage and other computer issues.
