Dismiss
InnovationQ will be updated on Sunday, Oct. 22, from 10am ET - noon. You may experience brief service interruptions during that time.
Browse Prior Art Database

Optimum Physical I/O Block Size Within a Given Merge Order for Random Data

IP.com Disclosure Number: IPCOM000079305D
Original Publication Date: 1973-Jun-01
Included in the Prior Art Database: 2005-Feb-26
Document File: 3 page(s) / 45K

Publishing Venue

IBM

Related People

Edel, TR: AUTHOR

Abstract

The main objective when handling data records is to minimize the time spent both reading and writing each record, so that the total time of processing all records in a file is as short as possible. Prior optimizations considered the problem solved if both reads and writes were fully overlapped. Full buffering, however, meant smaller physical block sizes. Thus, each logical record within a given physical block must be apportioned the start/stop time associated with physical data blocks. For tape operations the start/stop time is known as Inter-Record Gap time. For Direct Access Storage Device (DASD) operations, the start/stop time is known as the sum of rotational delay time plus seek time.

This text was extracted from a PDF file.
At least one non-text object (such as an image or picture) has been suppressed.
This is the abbreviated version, containing approximately 55% of the total text.

Page 1 of 3

Optimum Physical I/O Block Size Within a Given Merge Order for Random Data

The main objective when handling data records is to minimize the time spent both reading and writing each record, so that the total time of processing all records in a file is as short as possible. Prior optimizations considered the problem solved if both reads and writes were fully overlapped. Full buffering, however, meant smaller physical block sizes. Thus, each logical record within a given physical block must be apportioned the start/stop time associated with physical data blocks. For tape operations the start/stop time is known as Inter- Record Gap time. For Direct Access Storage Device (DASD) operations, the start/stop time is known as the sum of rotational delay time plus seek time.

The problem then is to combine both variables--Buffering and Block Size in an equation which gives the smallest transfer time per record.

(Image Omitted)

Note that iterations of T should be substituted in Equation 2 for values of N=0->M. When N=M+1 the average transfer rate is simply 2T and the AT equation does not apply. The lowest AT value obtained, thus dictates the number of overlap buffers (2(M+1)-N) which will give the smallest transfer time per logical record.

The ratio M-N over M is of particular note, since it represents the fraction of T which is completely I/O overlapped. Therefore, the Read and Write time of a single record is simply T. The value (1-(M-N over M)) represents the fraction of T that is partially buffered. That is: the Write time thus becomes (T/2 + T), since the Write is always double buffered until N=M+1. T being the normal write time, T/2 being the portion of the partially buffered read accrued to the next write.

Since the previous discussion is of great importance, the following example is shown:

(Image Omitted)

At any given point Time, data buffer D is being written while a simultaneous read occurs in any input buffer A, B, or C. (Example shows Buffer C being filled.) Note that the input b...