Maximizing Your RAID Investment
Everyone wants to get the best value when it comes to technology investments. In the case of RAID systems, that means planning for everything from monitoring to connectivity.
We buy insurance to protect our personal property, install smoke alarms to protect our homes, and perform routine maintenance on our cars to prevent breakdowns. We wouldn't even buy a television without considering all the options. According to Dieter Paul, president of StorCase (Fountain Valley, CA), and Steve Garceau, RAID product manager for CMD (Irvine, CA), technology investments deserve the same kind of consideration.
Having the right environment for your system is the first consideration. Even physical placement of the RAID unit can affect performance. Any vibration or movement slows down data access. However, heat is the number one enemy of RAID arrays. That means that every effort must be made to reduce and monitor temperature to extend the life span of the disks. Lower voltage controllers and cooler port components should also be employed. Options such as temperature sensors on every slot can be invaluable in preventing heat damage.
Since most IT professionals are too busy to spend their days in the server room, remote monitoring is a must. "You have to be able to tell when things are happening," says Paul. Vendors of various RAID components are offering tools to monitor RAID functions such as voltage and air flow. Smart interfaces which provide monitoring functions are available on many models, but, says Paul, the overhead and IT expertise needed may make them impractical. Relaying information through a separate bus to the Internet is easier and fosters collaboration. Remote monitoring can also track maintenance issues. For example, some programs display the number of hours a fan has been running. When that fan reaches the standard life expectancy, it can be replaced before a problem occurs. For VARs, this is an opportunity to enhance customer service. From anywhere in the world, a reseller could call a technician to schedule proactive maintenance. A good system, states Paul, should provide "an overall feel for performance at all times." He anticipates that 2001 will bring more monitoring options such as enclosures which communicate directly with pagers and fax machines.
Garceau points to advances in the use of SNMP (simple network management protocol) as a monitoring tool. By incorporating current Ethernet infrastructure, developers can monitor all components as well as configure remotely.
Always have a backup plan, advise Garceau and Paul. "For years people seemed content throwing in an extra disk drive, while relying on a single power source or controller," commented Garceau. No system should have a single point of failure at any stage in the process. For instance, dual controllers allow continued access to data even if one should go down. In the event of failure, the faulty controller could be hot-swapped. Having both actively processing enhances daily performance. Paul suggests maintaining a second RAID array at a remote site is an excellent "insurance policy." In the event of a fire or other catastrophe at the main site, the off-site RAID would still protect valuable data. When you consider the value of the data being stored, such precautions don't seem quite as expensive.
SCSI Or FC?
"Eventually everything will go with Fibre Channel," says Paul. Though many people think a RAID system can't go down, Paul warns there's always a slim possibility something can go wrong. Fibre Channel can provide the reliable connectivity to decrease the likelihood of failure. However, he emphasizes, SCSI is still the main type of connectivity and valuable for its relative simplicity. "When it comes to bare I/O, SCSI is still faster, especially with the introduction of Ultra 320." Any RAID investment at this point should be field upgradable to Fibre Channel or other emerging technologies.
Garceau also recognizes the industry's commitment to SCSI. The advantage for many integrators is the common design and upgrade capabilities. He feels that budgetary restrictions, lack of infrastructure, and the complexity of Fibre Channel installations may make some integrators hesitate. The industry's failure to promote a unified standard has also discouraged customers, in his opinion.
While Garceau acknowledges the long-term viabilty of SCSI and Fibre Channel, he is convinced that for many gigabit Ethernet will provide the connectivity answer. This is especially true in light of the popularity of SANs (storage area networks). With the Ethernet infrastructure already in place for networking and the existing expertise among IT professionals, Garceau feels that gigabit Ethernet is a logical step. Ethernet-attached storage products began hitting the market at the end of last year. "The good thing about this trend is that the industry has been doing all the preparation for this time," said Garceau.
With proper care, current RAID systems have a very long life span. Unfortunately, no investment in technology products is likely to be the final one. Even the best system will eventually need to be replaced as technology and productivity needs evolve. Perhaps the best option is to view a RAID system as a work in progress and plan accordingly.Questions about this article? E-mail the author at JackieM@corrypub.com.