|
Amidst
the unabated advance of faster computer hardware, the explosive growth of
storage, and the advent of virtualized storage becoming more commonplace
within the commercial setting, the performance of storage I/O has
undergone even greater scrutiny. Yet
surprisingly, the basic ability to readily assess (specifically, measure
and monitor) the performance of storage I/O operations at the file
level has been neglected. This
absence is particularly astonishing given that much (if not most) of the
actual work being performed by (faster) computer systems requires
significant file I/O activity.
We
contend that file I/O performance monitoring is a matter of practical
necessity and growing importance. We
also propose and briefly describe hIOmon™.
Featuring a flexible, adaptable and scalable architecture, this
novel, currently available implementation of a file I/O performance
monitoring facility affords a multitude of benefits to a wide audience
(including researchers, developers, vendors and computer end-users alike).
These benefits include salient improvements in the ability to
obtain empirical I/O performance data for better design decisions; better
diagnose and understand storage access performance problems; evaluate
proposed improvements to the performance of computer systems; verify and
ensure that required levels of file I/O performance (QoS) are being met;
help reduce storage management costs, and prudently approach and adopt
emerging storage technologies. |
