shouldnt I see IO improvement for installed image? I did a test and ran an image without installing it first, then I ran it after I installed it open/head/shared, but the installed image did more IOs in one activation than the uninstalled one ..any ideas why ?

> mc decw$clock.
> now run accounting
ACCOUNTING/TYPE=IMAGE/REPORT=(PROCESSOR,ELAPSED,DIRECT_IO,FAULTS,RECORDS)/SINCE=3-JUN-2008:10:30:00/SUMMARY=IMAGE

From: 3-JUN-2008 10:30 VMS Accounting Report To: 3-JUN-2008 10:35
Image name Processor Elapsed Direct Page Total
Time Time I/O Faults Records
--------------
DECW$CLOCK 0 00:00:00.31 0 00:00:03.61 47 389 1

> now install the image as shared,header resident.
install add sys$system:decw$clock.exe/open/header/shared/accounting
> run decw$clock
mc decw$clock
> run accounting again
ACCOUNTING/TYPE=IMAGE/REPORT=(PROCESSOR,ELAPSED,DIRECT_IO,FAULTS,RECORDS)/SINCE=3-JUN-2008:10:40:00/SUMMARY=IMAGE

From: 3-JUN-2008 11:40 VMS Accounting Report To: 3-JUN-2008 11:46
Image name Processor Elapsed Direct Page Total
Time Time I/O Faults Records
--------------
DECW$CLOCK 0 00:00:00.36 0 00:00:03.02 56 366 1

Since no experts is answering (maybe the HP ITRC forums would be a better place to ask :-), a few remark:
DECW$CLOCK single invocations do not proof much. It is interactive, i.e. not running the same time until stopped by You. And I don't know what counts as direct I/O and buffered I/O here. What You see is a reduction of page-faults in the installed case.

What really counts is the speed up in image loading, because directory lookup, file opening and (with /shared) allocating and mapping of shareable/shared sections is avoided; and if many shareable image sections are involved, paging at image startup can be reduced.

So installing makes sense for frequently used images, and/or images used by many processes.
And the effect of /shared very much depends on the number of shareable image-sections in the program.