Database Reference
In-Depth Information
•
Expected frequencies of individual queries
•
Anticipated frequencies of individual transactions
•
Time constraints
•
Uniqueness constraints in database tables
•
Indexes for each tables
•
Validity of primary and foreign keys
Use of Benchmarks
As the number of users increases, the size of the database becomes larger, the scope
of the applications expands, and the complexity of database operations intensifies,
the DBMS tends to become inadequate to provide acceptable performance. Queries
and transactions slow down. Conflicts among concurrent transactions escalate,
resulting in locking problems. Routine maintenance becomes more and more time-
consuming and difficult. You may begin to consider upgrading the DBMS or,
perhaps, even replacing it with a more robust one.
How do you go about making decisions to upgrade or replace the existing
DBMS? Performance benchmarks provide metrics that can be used for making
upgrade or replace decisions. You may apply benchmark tests to the existing DBMS
to check how inadequate it has become in the current environment. Also, bench-
mark tests can be used to justify upgraded versions or new DBMSs considered as
replacement.
Benchmarks quantify performance. They allow for variations in database envi-
ronments. Individual tasks cannot measure performance satisfactorily. You need a
collection of tasks to be used to quantify performance. Such a collection or suite of
tasks constitutes a performance benchmark. Typically, benchmarks measure per-
formance as transactions per second (tps). Also, they tie in with costs and compute
price/performance ratios ($/tps).
Benchmarks defined by the Transaction Processing Council (TPC) are well
known and are used more than some other benchmarks from industry. Other pro-
prietary benchmarks proposed by individual vendors are not very useful because
they are not universal. Let us examine the TPC benchmarks.
TPC-A.
First in a series of benchmarks. Defined in 1989, it measures performance
(tps and $/tps) in typical OLTP or operational environments. Simulates a standard
bank application with a single type of transaction similar to cash withdrawal and
deposit at a teller station, updating three tables, and adding an audit trail record in
another table. Measures end-to-end performance of the system including perfor-
mance at the workstation and on the data communications devices.
TPC-B.
Similar to TPC-A, but it measures only the core performance of the DBMS
and the OS by themselves. Not a measure for end-to-end performance. Used less
widely than TPC-A.
TPC-C.
More complex set of database operations than TPC-A or TPC-B. Based on
the model of an order entry transaction performing operations to enter order,
deliver order, record payment, monitor stock levels, and verify order status.
Search WWH ::
Custom Search