DocumentCode
2290376
Title
Empirical validation of availability models for the RISC System/6000 workstation using survey and measurement data
Author
Chandra, Arun ; Ahrens, George ; Kanthanathan, Mano ; Grzinich, John C.
Author_Institution
IBM Corp., Austin, TX, USA
fYear
1995
fDate
16-19 Jan 1995
Firstpage
439
Lastpage
444
Abstract
We show an empirically driven process to validate the availability models of RISC System/6000 workstations. The empirical data is obtained by tracking RISC System/6000 workstations in the field. The tracking data is obtained by using customer surveys and by using an availability measurement tool installed on selected running systems. We model several types of RISC System/6000 workstations using the SAVE tool. Care is taken to include all classes of systems. The workstations modeled include representatives from the desktop, deskside, and rack families. We explain the key principle of the “representative field model.” For each type of workstation modeled, availability data for validation is extracted from the customer survey database or the availability measurement database. The validation data is used to identify and eliminate model errors. The identification and elimination of modeling errors involves sensitivity analysis. The validation of the availability measures of several RISC System/6000 workstations using both survey and measurement data confirms the accuracy of our modeling process and gives us the confidence to utilize the availability models in our development and marketing processes. Also, this enables us to generate accurate availability measures in our future availability modeling efforts
Keywords
computer architecture; performance evaluation; reduced instruction set computing; workstations; RISC System/6000 workstation; SAVE tool; availability measurement database; availability measurement tool; availability models; customer survey database; customer surveys; data validation; marketing; measurement data; modeling errors; sensitivity analysis; survey data; Availability; Data mining; Databases; Power system modeling; Predictive models; Reduced instruction set computing; Sensitivity analysis; Software tools; Steady-state; Workstations;
fLanguage
English
Publisher
ieee
Conference_Titel
Reliability and Maintainability Symposium, 1995. Proceedings., Annual
Conference_Location
Washington, DC
ISSN
0149-144X
Print_ISBN
0-7803-2470-6
Type
conf
DOI
10.1109/RAMS.1995.513281
Filename
513281
Link To Document