home
***
CD-ROM
|
disk
|
FTP
|
other
***
search
/
Libris Britannia 4
/
science library(b).zip
/
science library(b)
/
PROGRAMM
/
DB_CLIPP
/
2614.ZIP
/
DB-TEST.ZIP
/
DB-TEST.DOC
next >
Wrap
Text File
|
1988-04-08
|
9KB
|
204 lines
DB-TEST.EXE
Version 1.0
By SCOTT TAYLOR
228 Gold Kettle Drive
Gaithersburg,MD 20878
DB-TEST is a test program to evaluate the time taken to seek
records in a database file. I created this test to determine how
different operating systems handled database performance. I
wanted to be able to compare relative times for different
operating systems, different hardware, and stand alone
workstations that are not on a network. The test results will
display each record as it is being accessed, and it will display
the total elapsed time in seconds required to perform the seeks.
I have found that different operatinng systems running on
identical hardware resulted in different seek times. This can
make a big difference on networks running multiuser database
applications. This test may be used on any PC running DOS 3.0 or
greater. If you will use this program on a standalone PC you must
run the DOS "SHARE.EXE" program in order to provide record
locking support. On some versions of DOS the share program is
called "SHARE.COM". If you use this program on a DOS compatible
network you must also have the "SHARE.EXE" program installed on
the workstations being used. The "SHARE.EXE" program is not
required for Novell networks since the DOS record locking support
is provided by the workstation shell. All stations running in the
test must have full access rights to the same root directory in
order to use the multi-user test mode.
When the test is being run the following takes place in the order
shown below:
1) Seek To The Desired Record
2) Lock That Record
3) Read The Data In That Record
4) Unlock That Record
5) Locate The Next Record And Repeat Above Steps
This will continue until all desired records have been accessed.
The results will show the total elapsed time (in seconds)
required to seek & read all records.
When this program is run it checks to see if the test database
file, TESTFILE.DB, and the database configuration file,
DB_QUEUE.CFG, exists in the root directory. If they do not exist
a message will appear indicating that a test database file must
first be generated. Before the test database file is generated
you will need to establish the record size and the number of
records to be used. If no test database has pervoiusly been
established a default configuration size will be displayed. The
default the size for each record is 1024 bytes (1K), and the
default number of records is 1000. This represents a total file
size of 1,024,000 bytes (or 1 megabyte). You must create a test
database using the displayed defaults or choose your own record
information.
Using a larger database size will create more work for the
operating system because record seeks will be further apart.
Using small record sizes will mean that seeks will occur more
frequently because the records are closer together and a smaller
amount of data will be read with each seek.
Using larger record sizes will cause seeks to occur less
frequently. It will be harder on the operating system and hard
disk because more data will be read with each seek and each
record is spread further apart.
The number of seeks to be performed during the test period must
be equal to or less than the total number of records in the
database file. If you select a seek style that uses a "SKIP"
pattern, the number of seeks will automatically be adjusted to
the maximum allowable value. For example if the test database
file has 1000 records in it and you select a seek pattern of
"SKIP BY 10", the number of seeks to be performed will be
adjusted to 100. This is because the test will skip over 10
records per seek, therefore 1000 records divided by 10 skips is
100 seeks.
The seek style is the pattern of seeking that you want to be
performed during testing. The different patterns are described
below in the menu options section.
For comparing operating systems you should use the same type CPU
and speed, as well as the same hard disk. Then you would only
change operating systems. After running the tests on each you can
compare the relative results.
STATISTICS:
The statistics window displays the results for the test in
progress. The following details the results displayed.
CURRENT RECORD: This displays the actual record number that
is currently being read.
CURRENT SEEK: This displays the current seek being
performed. This value is incremented after each seek until
the total number of seeks specified have been executed.
ELAPSED SECONDS: This will display the total number of
seconds that have elapsed since the test was started. This
value is updated after every record seek. It is displayed
with an accuracy of 1/10 seconds.
MENU OPTIONS:
The right and left arrow keys are used to highlight the
different menu options. Pressing the return key allows you
to enter a value for that option. The <F1> key is used to
start testing in a single station mode. The <F2> key is used
to activate the multi-user mode. After all stations have
been activated into the multi-user mode you should press the
<F1> key at the last workstation and each node will begin
testing simultaneously. The <F3> key is used to generate a
test database file using the RECORD SIZE & NUMBER OF RECORDS
shown in the menu. A database file must exist before
testing. The <ESC> key is used to exit the program to DOS.
Instructional help text is displayed for each menu option
that it is selected.
RECORD SIZE: This is the size of each record contained in
the test database. If a test database file already exists
the current configuration of that file will be displayed.
If no database file exists than a default of 1024 bytes is
displayed. The range allowed is from 128 bytes to 8192 bytes
(8K bytes). A larger record size will increase the
throughput and will generate a larger test database file
size. Simply type a value within the allowable range and
press return to accept.
NUMBER OF RECORDS: This is the total number of records that
exist in the test database. If a test database file exists
then the current configuration will be displayed. If no test
file exists then a default of 1000 records will be
displayed. The range allowed is 1 record to 10000 records.
More records will generate a larger test database file and
the larger the test file the better the relative results
will be. Simply type a value in the allowable range and
press return to accept.
NUMBER OF SEEKS: This is the total number of seeks you wish
to be performed for the current test period. The number of
seeks cannot be larger that the total number of records
allocated in the test database file. The range is from 1 to
10000 seeks, or the maximum number of records. The maximum
will be displayed in the help test. Just type in a value
within the allowable range and press return to accept.
SEEK STYLE: This is the seek pattern to be used in testing.
Use the right arrow key to select a different seek style, or
pattern, then press return to accept. The different patterns
are as follows:
1) OUTER TO INNER
This would seek records from the outer portions of
the test database file and work inward. An example
would be to seek to records 1, 1000, 2, 999, 3, 998,
4, 997, 5, 996, etc...
2) INNER TO OUTER
This is the opposite of OUTER TO INNER. It would
seek
records from the middle of the database file
outward. An example would be to seek to records
500, 501, 499, 502, 498, 503, 497, 504, etc...
3) SKIP BY 10
This would start seek at the beginning of the test
file and skip over 10 records at a time. An example
would be to seek to records 1, 10, 20, 30, 40, 50,
60, etc...
4) SKIP BY 100
This would start seeking at the beginning of the
test database file and skip over 100 records at a
time. An example would be to seek to records 1, 100,
200, 300, 400, 500, 600, etc...
5) RANDOM SEEKING
This would seek to randomly selected records in the
database test file.