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

Time-Stamp Synchronization for Parallel Trace Services

IP.com Disclosure Number: IPCOM000105098D
Original Publication Date: 1993-Jun-01
Included in the Prior Art Database: 2005-Mar-19
Document File: 2 page(s) / 63K

Publishing Venue

IBM

Related People

Bucher, CK: AUTHOR [+3]

Abstract

Disclosed is a low-cost and sufficient method to synchronize the time stamps of the trace records that are generated from different computers in a shared nothing distributed parallel system.

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

Time-Stamp Synchronization for Parallel Trace Services

      Disclosed is a low-cost and sufficient method to synchronize
the time stamps of the trace records that are generated from
different computers in a shared nothing distributed parallel system.

      A shared-nothing distributed parallel system is composed of
multiple computers that are linked together with communication
channels.  A user's job can be executed by multiple processes in
several different computers in the shared-nothing distributed
parallel system.  In other words, a trace report of a user's job can
be generated from multiple trace files that are produced on different
computers.  However, the system clock of each computer in the
distributed parallel system usually is not synchronized.  This makes
it very difficult to sort the trace records of the distributed
parallel system by their time stamps.  A common clock among the
computers in the distributed parallel system is required to generate
an accurate trace report.

      In a distributed parallel system, the computer that accepts the
user's job requests is called a coordinate computer.  The computer
that processes the user's job is called a subordinate computer.  The
system clock of the coordinate computer is chosen as the common clock
of the distributed parallel system.  The time difference between the
common clock and the clock of each subordinate computer are collected
and logged into a file called the system time trace file at the
system start-up time by the coordinate computer.  The time difference
record contains the coordinate computer ID, the subordinate computer
ID, the time offset between the common clock and the clock of the
subordinate computer, and the system time of the subordinate comp...