| If you would like a more detailed number, I suggest that you use VPA (or
even VMS ACCOUNTING).
With VPA (now called DECps) you identify all of the Oracle images. (Obviously,
you need some one with Oracle knowledge to help identify them or you can guess
at the names via the directory locations.)
You then create a VPA workload family that contains these images. Call it
ORACLE or whatever. From the workload, you can now generate all sorts of VPA
graphs and reports, including your CPU graph.
It worked for me. If you don't have VPA, you can basically do the same
with VMS ACCOUNTING (with image accounting turned on), but you'll have to do
all of the work putting the graphs together by yourself.
All of this helps you get CPU usage, not response time.
See next reply for response time.
Steve
|
| In a traditional mainframe environment (ex: IBM MVS/ESA or Burroughs MCP/AS),
response time/service time is often defined as the elapsed time from when the
Enter or transmit key was pressed until the screen is resdisplayed. Usually,
the response time can be measured by the teleprocessing monitor (CICS or GEMCOS)
or by an external entity such as the communications system (ex: VTAM or the NSP)
which can record the time that the input data was received and the time that the
output screen was sent.
Digital's communications subsystem is not block mode and has no knowledge of
when a screen has been "transmitted to the host". (Actually, we don't even
transmit the screen to the host, it's already there.) Because of this, response
time can only be measured by/within the application or the application system.
i.e. If Oracle's SQL*FORMS or the ORACLE database doesn't track the response
time with some built in functions, and the customer's didn't put it in the
application, you probably won't be able to get a meaningful number. VPA does
have a reponse time facility, but it still won't really provide the kind of
information that your customer PROBABLY expects.
IMHO, the concept of response time for a screen breaks down with the current
trend to interactive computing. How meaningful is that number when the
program/DECFORM/4GL/etc... immediately validates the contents of a field as
soon as I finish the field (not the form)? In a block mode environment, you only
do this once when the form is received. In an interactive environment, you may
validate fields/pop up a list of values/go to the database several times during
the course of filling out a form! This really trashes the tradtional concept
and measurement of response time. Nonetheless, customers continue to ask for
it...
Steve
|