Browse Prior Art Database

Dynamic Compression During System Save Operations

IP.com Disclosure Number: IPCOM000042274D
Original Publication Date: 1984-May-01
Included in the Prior Art Database: 2005-Feb-03
Document File: 2 page(s) / 14K

Publishing Venue

IBM

Related People

Crowley, TR: AUTHOR [+3]

Abstract

In a computer system it is necessary to save the data base in a reasonable amount of time. The time needed to save the data base depends upon the speed of the central processing unit (CPU), the data transfer rate of the direct storage access device (DASD) and the data transfer rate of the save device. The save procedure designed with these factors considered often performs well in one environment but performs poorly in another. To solve this problem, the save procedure synchronizes its components to achieve the optimal save process. In most cases, the data transfer rate of the DASD is significantly higher than the data transfer rate of the save device. This implies that the read task (Fig. 1) can transfer data much faster than the write task (Fig. 2) can write the data to the save media.

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

Page 1 of 2

Dynamic Compression During System Save Operations

In a computer system it is necessary to save the data base in a reasonable amount of time. The time needed to save the data base depends upon the speed of the central processing unit (CPU), the data transfer rate of the direct storage access device (DASD) and the data transfer rate of the save device. The save procedure designed with these factors considered often performs well in one environment but performs poorly in another. To solve this problem, the save procedure synchronizes its components to achieve the optimal save process. In most cases, the data transfer rate of the DASD is significantly higher than the data transfer rate of the save device. This implies that the read task (Fig. 1) can transfer data much faster than the write task (Fig. 2) can write the data to the save media. The speed of the save device is the restraining factor in the save operation. This condition is referred to as "I/O bound". Some form of data compression is implemented to compensate for the difference between the data transfer rates of DASD and the save media. The data compression reduces the amount of data that must be written to the save media and, thus, optimizes the write task. However, implementing a compression algorithm which reduces the amount of data to be written to the save media will not necessarily optimize the save operation. Many different compression algorithms exist, but it is generally true that the higher the degree to which the data is to be compressed, the more time is required to compress the data. When maximum compression is used, the write task will be forced to wait for the compress task (Fig. 3) to compress the buffer before it can be written to the save media. In this case, the speed of the CPU becomes the restraining factor in the save operation. This condition is referred to as "CPU bound". The CPU and I/O resources must be balanced in order to achieve the optimal save procedure. The various parameters, such as CPU speed, DASD data transfer rates, and save device data transfer rates, are used to determine the degree of compression which should be implemented to balance the tasks. If these decisions are made during the design phase, the save procedure is locked into a compression technique which performs adequately for the majority of its users. However, compression techniques tend to be dependent on the data being compressed and some users will experience performance degradation because of the nature of their data. Also, to be flexible to changes, the compression algorithm should depend on the three parameters previously mentioned. The following system configuration changes would affect the parameters: - A new model of the CPU. - A different DASD attached to the system. - A different save device attached to the system. As these parameters change, the balancing of the tasks may be adversely affected. This problem could be partially solved by a configuration table...