Block and volume management doesn't cut it anymore
By Linda Christie
A recent survey of more than 430 enterprise data centers and leading storage service providers conducted by ITCentrix , a developer of IT decision support solutions for CIOs,
As many disk storage makers migrate from 36GB hard disks to 72GB and even larger spindles inside arrays, the likelihood of performance problems due to disk contention increases dramatically. "When multiple users access a spindle simultaneously, they must share the bandwidth of a single read-write channel," said Chris Gahagan, V.P. and General Manager, Recovery and Storage Management for BMC Software . "As spindle density doubles, the chances of I/O collision potentially doubles, too -- even with higher speeds. So data placement is becoming a key issue for improving efficiency."
To stabilize responsiveness, Gahagan says data centers will need to map applications down to the lowest common denominator: the I/O head inside the disk drive. "With Oracle Parallel Server, for example, many data files are accessed simultaneously. So, if two high performance tables reside on the same spindle, your customers will experience delays -- not a pretty picture if they're trying to view their bank statements or place orders on your website. Indexes, especially, need to reside on separate platters to avoid collision."
A similar, related problem occurs when a low-priority application contends for storage with a high priority application; in this case, the more important application suffers because of poor data placement.
Mapping applications becomes even more complex when you consider that businesses run in cycles causing peaks and valleys of applications usage. "The level of spindle access isn't static," Gahagan said. "Managing data placement is a dynamic process."
Unfortunately storage technology is making it harder to pinpoint this information. "Array vendors don't understand applications and their use of storage," Gahagan said. "So, you'll need tools that automate this process."
BMC Software's advanced data management solution, including PATROL(R) and Application-Centric Storage Management(TM) for example, tracks the performance of an Oracle database all the way down to the spindle level. Graphic presentations and reports of actual running databases reveal what SQL statements are most actively utilized as well as how tablespaces are performing, so you can make adjustments.
"With the Internet, your infrastructure is visible to the public now," said Gahagan. "So, responsiveness has become a critical factor for ensuring customer confidence."
About the author: Linda Gail Christie is a contributing editor based in Tulsa, Ok.
* Check out searchStorage Best Web Links for other storage management resources.
* What are your thoughts on automated storage management? Post these and other storage management issues in our Storage Management Tips and Tricks discussion forum.
* For other topics of interest, check out our archive of Storage Management tips.
This was first published in May 2001