User Guide

64
Chapter 4
Ratio of computation to output. Commands that perform a lot of computation and
produce small output results (for example, few and small pivot tables, brief text
results, or
few and simple charts) have the most to gain from running in distributed
mode. The degree of improvement depends largely on the computing power of
theremoteserver.
Small jobs.
Jobs that run quickly in local mode will almost always run slower in
distributed mode because of inherent client/server overhead.
Charts. Case-oriented charts, such as scatterplots, regression residual plots, and
sequence c
harts, require raw data on your local computer. For large data files or
database tables, this can result in slower performance in distributed mode because the
data have to be sent from the remote server to your local computer. Other charts are
based on s
ummarized or aggregated data and should perform adequately because the
aggregation is performed on the server.
Interactive graphics. Sinceitispossibletosaverawdatawithinteractivegraphics
(an optio
nal setting), this can result in large amounts of data being transferred from
the remote server to your local computer, significantly increasing the time it takes to
save your results.
Pivot ta
bles.
Large pivot tables may take longer to create in distributed mode. This is
particularly true for the OLAP Cubes procedure and tables that contain individual
case data, such as those available in the Summarize procedure.
Text out
put.
The more text that is produced, the slower it will be in distributed mode
because this text is produced on the remote server and copied to your local computer
for display. Text results have low overhead, however, and tend to transmit quickly.
Server Login
The Server Login dialog box allows you to select the computer that processes
commands and runs procedures. This can be either your local computer or a remote
server
.