TPC-D: The Challenges, Issues and Results.
Ramesh Bhashyam:
TPC-D: The Challenges, Issues and Results.
VLDB 1996: 593@inproceedings{DBLP:conf/vldb/Bhashyam96,
author = {Ramesh Bhashyam},
editor = {T. M. Vijayaraman and
Alejandro P. Buchmann and
C. Mohan and
Nandlal L. Sarda},
title = {TPC-D: The Challenges, Issues and Results},
booktitle = {VLDB'96, Proceedings of 22th International Conference on Very
Large Data Bases, September 3-6, 1996, Mumbai (Bombay), India},
publisher = {Morgan Kaufmann},
year = {1996},
isbn = {1-55860-382-4},
pages = {593},
ee = {db/conf/vldb/Bhashyam96.html},
crossref = {DBLP:conf/vldb/96},
bibsource = {DBLP, http://dblp.uni-trier.de}
}
Abstract
This presentation covers what we at NCR have learned
about the TPC-D benchmark as we executed and published our first
set of volume points for the Teradata Database. Areas where customers
should read the Full Disclosure Report carefully are pointed out
as well as the weaknesses in the benchmark relative to real customer
applications. The key execution and optimization elements of the
Teradata Database and the 5100 WorldMark platform that contribute
to our published results are discussed.
The Benchmark
The TPC-D benchmark models a decision support environment through
a collection of 17 complex queries and two update functions. It
also specifies other necessary issues such as the transaction
ACID properties, and database maintenance options. The benchmark
specifies six database scale factors- 1 GB, 10 GB, 30GB, 100GB,
300GB, 1000GB, and the results are reported as three primary metrics:
the power metric representing the single user query per hour rate,
the throughput metric which represents the multiple user throughput
capacity, and a price/performance metric which is very important
for the MPP systems typically used for the VLDB database sizes.
Going Behind The Metrics
It is however necessary to understand a lot more about the benchmark
than just the numbers. The benchmark requires both an executive
summary report and a full disclosure. Much of this presentation
focuses on the items in these reports which are necessary in order
to understand the performance characteristics of a particular
product.
These items are: the individual run times of the queries, the
index structures used, the data to disk ratio, the number of concurrent
query streams, the level of data protection, the methods of database
maintenance, the data load time, and the full SQL syntax used
to execute the queries.
Benchmark Improvements
We present three suggestions for improvements to the benchmark.
These suggestions relate to disclosure of items related to the
execution and reporting of the benchmark which we feel are important,
particularly to the VLDB customer, in helping to relate the benchmark
to the real world.
Key Teradata DBS Features
We also present the key features of the Teradata DBS which contribute
to our published results. These features include: fully parallel
DBS execution, throughput mechanisms, query optimizations, enhanced
evaluation of rows, efficient join plans, and global query optimizations.
Each of these is presented and examined for its contribution
to query performance at VLDB volume points.
Conclusion
The decision support area is complex and the TPC-D benchmark query
suite is a good attempt at capturing this complexity. Because
of this complexity, it is necessary to look beyond the primary
metric numbers to the details of the performance, the database
and the environment in which it was executed. Only with this can
one judge the degree to which a published instance of the TPC-D
benchmark applies to the customer's application and workload.
The full text of a document and our latest TPCD results
are available at:
http://www.ncr.com/product/teradata
Copyright © 1996 by the VLDB Endowment.
Permission to copy without fee all or part of this material is granted provided that the copies are not made or
distributed for direct commercial advantage, the VLDB
copyright notice and the title of the publication and
its date appear, and notice is given that copying
is by the permission of the Very Large Data Base
Endowment. To copy otherwise, or to republish, requires
a fee and/or special permission from the Endowment.
Online Paper
CDROM Version: Load the CDROM "Volume 1 Issue 5, VLDB '89-'97" and ...
DVD Version: Load ACM SIGMOD Anthology DVD 1" and ...
Printed Edition
T. M. Vijayaraman, Alejandro P. Buchmann, C. Mohan, Nandlal L. Sarda (Eds.):
VLDB'96, Proceedings of 22th International Conference on Very Large Data Bases, September 3-6, 1996, Mumbai (Bombay), India.
Morgan Kaufmann 1996, ISBN 1-55860-382-4
Contents 
Electronic Edtion
Copyright © Tue Mar 16 02:22:06 2010
by Michael Ley (ley@uni-trier.de)
