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

Manipulation of Hot Spots in Distributed Databases

IP.com Disclosure Number: IPCOM000120661D
Original Publication Date: 1991-May-01
Included in the Prior Art Database: 2005-Apr-02
Document File: 3 page(s) / 158K

Publishing Venue

IBM

Related People

Thomasian, A: AUTHOR

Abstract

Disclosed is a method to distribute and manipulate data pertaining to the allocation of resources in a distributed database environment. The method is based on the rationing of resources in anticipation of demand at various nodes, such that requests are satisfiable locally most of the time. Resources are periodically replenished adapting to demand history and active database technology -1- is used to automatically re-adjust supplies.

This text was extracted from an ASCII text file.
This is the abbreviated version, containing approximately 40% of the total text.

Manipulation of Hot Spots in Distributed Databases

      Disclosed is a method to distribute and manipulate data
pertaining to the allocation of resources in a distributed database
environment.  The method is based on the rationing of resources in
anticipation of demand at various nodes, such that requests are
satisfiable locally most of the time. Resources are periodically
replenished adapting to demand history and active database technology
-1- is used to automatically re-adjust supplies.

      Aggregate objects, e.g., the quantity-on-hand (QOH) of a
frequently used inventory item are sources of lock contention in
centralized and distributed environments. Specialized locking modes,
such as increment/decrement locks, are one method to reduce the lock
contention on such objects when they become a hot spot in a
centralized system. In some cases it is important to monitor the
value of such a quantity so that it stays within an acceptable range.
For example, when an aggregate object, such as QOH, is decremented,
it is necessary to assure that the value remains positive.  This is
handled by field calls in IMS Fast-Path.  Briefly, the application
program checks whether QOH is greater than the requested quantity.
This check is repeated again before QOH is updated at transaction
commit time.  The transaction is aborted if the test fails the second
time around, while the programmer can usually provide an alternative
strategy when the test fails initially.

      The field calls method of IMS has been generalized to escrow
locking in -2-.  In escrow locking the user makes special calls to
indicate the delta by which the aggregate data item is to be
incremented/decremented.  Each aggregate quantity is represented by a
3-tuple (its minimum, current, and maximum value).  The current value
indicates the value of the data item reflecting transaction commits,
while the range (maximum-minimum) reflects the range of possible
values that the data item can attain when all in-progress
transactions are committed/aborted.  This is a more intelligent test
than field calls in IMS Fast-Path, in that if a transaction succeeds
an initial test it will never be denied at a later point.  A
generalization of this concept to the data (or disk) sharing
environments appears in -3-. A shortcoming of the above approach is
the necessity to introduce special calls for handling aggregate data,
which may require complicated extensions to the lock manager.

      Active databases technology is extended to a distributed
database environment to control the re-allocation of the data as
determined by demands for the object.  The scheme reduces
inter-system communication required to access high-traffic aggregate
objects in a distributed database environment by a new data
allocation and manipulation method, which is referred to as data
rationing.  The method is based on allocating fractions of an
aggregate quantity, such that the sum of the partitions equals t...