Browse Prior Art Database

Data Set Load-Balancing Algorithm

IP.com Disclosure Number: IPCOM000061493D
Original Publication Date: 1986-Aug-01
Included in the Prior Art Database: 2005-Mar-09
Document File: 1 page(s) / 12K

Publishing Venue

IBM

Related People

Henkels, RG: AUTHOR [+3]

Abstract

This algorithm represents a load-balancing mechanism for distributing write requests to a group of page data sets. A page data set provides temporary storage on the direct-access storage device (DASD) for pages cast out of system main storage to free-up page frames needed for more active pages. The load-balancing algorithm may place any data record on any of the page data sets in the group. This algorithm can balance the workload among all the available data sets in the group and is particularly useful when they are on mixed direct DASD configurations. The algorithm also allows for rapid reaction to changes in device performance due to contention or other anomalies. Each of the data sets in the group is evaluated in a "round-robin" order to determine whether it should be selected to receive additional work.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 51% of the total text.

Page 1 of 1

Data Set Load-Balancing Algorithm

This algorithm represents a load-balancing mechanism for distributing write requests to a group of page data sets. A page data set provides temporary storage on the direct-access storage device (DASD) for pages cast out of system main storage to free-up page frames needed for more active pages. The load- balancing algorithm may place any data record on any of the page data sets in the group. This algorithm can balance the workload among all the available data sets in the group and is particularly useful when they are on mixed direct DASD configurations. The algorithm also allows for rapid reaction to changes in device performance due to contention or other anomalies. Each of the data sets in the group is evaluated in a "round-robin" order to determine whether it should be selected to receive additional work. The algorithm computes the "load" on each data set in the group, and selects a data set only if the load is less than or equal to the average system load for all the data sets in the group. Therefore, data sets which are handling their workload better than average are selected more frequently than data sets which, for whatever reason, are not handling their load as efficiently. The computation of "load" for each data set is based on two factors: device service time per input/output (I/O) operation, and the number of I/O requests which are queued on the data set at the time the evaluation is made. In order for the data set selection algorithm to be computationally effective, the service times and requ...