IDE RAID provides low-cost option
It may be surprising, but small and midsize businesses still run network servers and workstations primarily on IDE storage technology. With the proliferation of NT client systems and larger and faster IDE drives, users and IS professionals depend increasingly on IDE drives to store critical project data. This trend will probably continue with the introduction of the DMA 66 standard.
That`s great, until a drive fails. When that happens, administrators and users learn the painful lesson of why last week`s backup is of limited value in the aftermath of today`s system crash.
Without some degree of fault-tolerant storage, you`re dealing not so much with "What if the system fails?" but "How much risk can we assume without fault tolerance?" For the shortsighted, that issue isn`t considered until a drive fails.
I know a colleague whose livelihood in marketing depended on a large database of current and potential customers. He got comfortable--overly so, it turns out--with a modest, dual-drive server`s reliability in managing his department`s database. When the main drive failed, the second took over. Trouble was, the failed drive`s data hadn`t been backed up to the second drive for more than a month, losing critical order, backlog, and credit data. He and his staff spent the next two weeks rebuilding that database--and explaining to management why a once-efficient department had been brought to its knees.
A hair-pulling episode like that can be avoided with a little forethought. But if, like me, your customers, staff, projects, and budget typically consume your thoughts, you don`t want to be in a position in which you`re preoccupied with your organization`s data integrity. Here are some solutions that should keep you covered.
The first is to set your application software to back up every few minutes. This typically is a single-drive situation, so it only offers minimum protection against corrupted files and crashes. Some application software features this capability.
Microsoft, for example, builds this feature into its Office products. With this approach, you`re forced to live with periodic interruptions of work while the program writes updates to your drive. At the very least, this can be a distraction, if not an ongoing disruption to your work, particularly for "power users." On top of that, these solutions still don`t secure data if the hard drive fails.
Until recently, installing a SCSI RAID system was the only fault-tolerant solution. But SCSI RAID is relatively expensive, even by today`s standards. Now, however, there are multiple alternatives to SCSI RAID, which use lower-cost IDE drives for writing to the primary and secondary drives simultaneously.
Windows NT, for example, includes software RAID capability, providing RAID 1 and 5. This is a good solution for light server loads, but it constantly uses processor cycle time. Performance loss is one thing, data loss is another. So, if software RAID is your only choice, buy another drive and use this safe, yet processor-intensive, approach.
A hardware/software approach is an excellent choice and provides relatively good performance. One example is RAID technology on a PCI card, which provides considerably more flexibility than a software-only solution, including RAID 0 and support for up to four drives.
The downside is that this approach requires the system bus to manage some functions and it uses drivers that are subject to quirks in operating system revisions or updates.
Hardware-only products are relatively new to IDE RAID, but from a performance standpoint they represent the fastest and easiest solution to implement. IDE hardware-only solutions provide a plug-and-forget approach. Because of their minimal overhead, they`re transparent to the computer, self-contained, and don`t interfere with overall system performance.
Any veteran IS professional or savvy user will tell you that, regardless of the method chosen, a fault-tolerant system is imperative to data protection. However, the scale of the computer network determines the scope of its fault tolerance. SCSI RAID is the answer for large-scale systems, albeit at a price premium. IDE RAID, on the other hand, is emerging as a simple and reliable answer for desktop, workstation, and small single-server environments.
George Zraick is vice president of sales and marketing at Connector Resources Unlimited Inc. (www.cruinc.com), Milpitas, CA.