Browse Prior Art Database

Return On Consumption-Based Dynamic Buffer Management Scheme In Query Environments

IP.com Disclosure Number: IPCOM000100641D
Original Publication Date: 1990-May-01
Included in the Prior Art Database: 2005-Mar-15
Document File: 2 page(s) / 70K

Publishing Venue

IBM

Related People

Cornell, D: AUTHOR [+2]

Abstract

A dynamic buffer management scheme is described to allocate storage among concurrently executing queries. It uses the concept of buffer consumption and return on consumption to guide storage allocation.

This text was extracted from an ASCII text file.
This is the abbreviated version, containing approximately 52% of the total text.

Return On Consumption-Based Dynamic Buffer Management Scheme In Query Environments

       A dynamic buffer management scheme is described to
allocate storage among concurrently executing queries.  It uses the
concept of buffer consumption and return on consumption to guide
storage allocation.

      The concepts of buffer consumption and return on consumption
(ROC) are introduced to guide storage allocation.  Buffer consumption
of a query is defined to be the integral of instantaneous buffer
allocation over the lifetime of the query.  A complex query can be
viewed as a sequence of steps where each step is either a two-way
join or some retrieve or projection operation.  Assuming the storage
allocation stays the same during the lifetime of a query step, the
buffer consumption of a query step can be expressed as the in-memory
time of a query step multiplied by the storage allocation.  Note that
the in-memory time here is from the time the query step is initiated
until it is completed.  It does not include the time waiting for
storage to become available.  ROC is the reduction in in-memory time
divided by the additional buffer consumption relative to the point of
minimum buffer allocation required to execute the query access plan.

      To determine how to allocate storage among concurrent executing
queries, we propose a simple dynamic scheme based on ROC.  Since in-
memory time of a query step is determined mainly by IO time, we can
use the stand-alone execution time of a query step as an estimate of
its in-memory time.  For each query step Qi, let qmax be its ROC at
the maximum allowed storage allocation constraint per query step or
the maximum requ...