Can you please elaborate a little on the parameters to the utility? (I am specifically
interested in S/M/L database size and think time.)

S/M/L refer to the size of the database—10 MB/1 GB/100 GB. Things like maximum
number of products and customers are tied to the size. Think time is the time between
requests, while the simulated user is reading the previous response and "thinking."
You can set it to a value you think is realistic (10 seconds?, 2 seconds?) and use this
value to estimate the number of total supported users. Or you can set it to 0 and have
the program drive the back end as hard as possible for a measurement of total
throughput (orders per minute). Or you can adjust it if you are trying to drive an
arbitrary number of orders per minute.


Is the number of threads equivalent to the number of users?

Number of threads is the number of connections to the back end (Web server or
database). If you set think time to a realistic value (as discussed in the previous
question) then threads will be equal to users.


What is the difference between starting the program with 10 threads from a single
machine compared to 5 threads from two different machines?

Unless you are hitting a resource bottleneck (either the client CPU or networking) you
shouldn't see much difference.


How do I interpret the data obtained from the ds2webdriver utility? (For example,
et = 100.0 n_overall = 280 opm = 168 rt_tot_lastn_max = 298 rt_tot_avg = 115
rollbacks: n = 121 % = 43.2.)

See ./ds2/drivers/ds2_driver.txt


Is it fair to compare the DS2 database size (10 MB/1 GB/100 GB) with the number
of warehouses in TPCC?

No


I get the following error while loading the DVD Store on SQL Server: Msg 4866,
Level 16, State 1, Line 1 The bulk load failed. The column is too long in the data file
for row 1, column 20. Verify that the field terminator and row terminator are
specified correctly.

This error is a result of the input load files being Linux/Unix based, while Microsoft
SQL Server is expecting DOS-based text files. The difference between the two is that Unix
text files just have a line feed, but DOS text files will have a carriage return at the end of
each line. As a result, the SQL load function never finds the carriage return and thinks
that the data is too big. If a version of Winzip is used to extract the ds2.tar file, then the
text load files are automatically converted. There are also many Unix2DOS utilities available
on the Web that can be used to do the conversion as well.

Return to DVD Store main page