Publication Details
Title: Measuring with Slow Clocks
Author: H. Beilner
Group: ICSI Technical Reports
Date: July 1988
PDF: http://www.icsi.berkeley.edu/pubs/techreports/tr-88-003.pdf
Overview:
This report describes a measurement technique and corresponding statistical evaluation options that can be used for assessing the mean duration of performing a particular operation, even when this duration is small compared with the resolution of an available, readable clock. The technique has been developed with regard to measuring operation durations of distributed system kernels, and to measuring durations of sub-activities embedded in these operations. The technique employs repetitive executions of the measured operation, but does not, however, depend on the usually employed "tight loop" around the operation. It also allows for simultaneous assessments of several different time intervals within the repetitive pattern. Based on an initial guess about the mean length of the smallest time interval to be measured, the necessary number of loop cycles can be determined before an experiment, for a selectable width of the confidence interval of the mean to be estimated, and at a selectable confidence level.
Bibliographic Information:
ICSI Technical Report TR-88-003
Bibliographic Reference:
H. Beilner. Measuring with Slow Clocks. ICSI Technical Report TR-88-003, July 1988
Author: H. Beilner
Group: ICSI Technical Reports
Date: July 1988
PDF: http://www.icsi.berkeley.edu/pubs/techreports/tr-88-003.pdf
Overview:
This report describes a measurement technique and corresponding statistical evaluation options that can be used for assessing the mean duration of performing a particular operation, even when this duration is small compared with the resolution of an available, readable clock. The technique has been developed with regard to measuring operation durations of distributed system kernels, and to measuring durations of sub-activities embedded in these operations. The technique employs repetitive executions of the measured operation, but does not, however, depend on the usually employed "tight loop" around the operation. It also allows for simultaneous assessments of several different time intervals within the repetitive pattern. Based on an initial guess about the mean length of the smallest time interval to be measured, the necessary number of loop cycles can be determined before an experiment, for a selectable width of the confidence interval of the mean to be estimated, and at a selectable confidence level.
Bibliographic Information:
ICSI Technical Report TR-88-003
Bibliographic Reference:
H. Beilner. Measuring with Slow Clocks. ICSI Technical Report TR-88-003, July 1988