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

Method to Obtain the Lock Request Workload Data without Paying the Collection Overhead

IP.com Disclosure Number: IPCOM000106679D
Original Publication Date: 1993-Dec-01
Included in the Prior Art Database: 2005-Mar-21
Document File: 2 page(s) / 62K

Publishing Venue

IBM

Related People

Shi, S: AUTHOR [+2]

Abstract

The incoming rate of lock requests for the OS/2* Database Manager is important performance information which can be used to tune the OS/2 Database Manager*. However, the collection of the rate of lock requests can impose a significant performance overhead on the OS/2 Database Manager. This disclosure provides a way to estimate the rate of lock requests without actually collecting it.

This text was extracted from an ASCII text file.
This is the abbreviated version, containing approximately 53% of the total text.

Method to Obtain the Lock Request Workload Data without Paying the Collection Overhead

      The incoming rate of lock requests for the OS/2* Database
Manager is important performance information which can be  used to
tune the OS/2 Database Manager*.  However, the collection of the rate
of lock requests can impose a significant performance overhead on the
OS/2 Database Manager.  This disclosure provides a way to estimate
the rate of lock requests without actually collecting it.

      The incoming  rate of lock requests for the OS/2 Database
Manager is an important performance metric.  It can be used to
determine whether there are too many lock escalations, and then
figure out whether the locklist size of the Database Manager should
be increased so that the number of lock escalations can be reduced to
increase concurrency.

      However, collecting this data can impose a significant overhead
over the OS/2 Database Manager performance.  Given that the OS/2
Database Manager can provide performance data like average response
time for unit of work and number of locks held in the system, a
simple queueing model can be used to estimate the rate of lock
requests from these two metrics.

      The Little's formula [*]  represents a simple queueing model:

Average queue length = (Average service time per request * Rate of
requests)

If this simple model is applied to the scenario, the following
formula results:

    Average number of locks held in the system =
    (Average service time per lock request * Rate of lock requests)

Hence,

Rate of lock requests = (Average number of locks held in the
system/Average service time per request)

Therefore,

    Rate of lock requests =
    (...