VMS 7.3-2 SYSv1800 paged pool lookaside lists

[SDIH1]

In the release notes of SYSv18 I found this:
...
5.1 New functionality addressed in this kit

5.1.1 Introduced Paged Pool Lookaside Lists

5.1.1.1 Functionality Description:

Paged pool lookaside lists are being introduced. This will benefit systems experiencing paged pool fragmentation and related performance slowdowns. This
feature will be off by default. To enable it, the SYSGEN parameter RSVD_EXEC_1 should be set to 512.
...

Obvious questions:
- I thought lookaside lists were already available, or was that limited tot nonpaged pool until now?
- How can I tell if our system would benefit from turning RSVD_EXEC_1 to 512, and how do I measure the difference?

[Pramod Kumar M · Accepted solution]

I thought lookaside lists were already available, or was that limited to nonpaged
pool until now?
Yes, it was limited to nonpaged pool until now. With the VMS732_SYS-V1800 paged
pool also starts using look aside lists if SYSGEN parameter RSVD_EXEC_1 is set
to 512.

How can I tell if our system would benefit from turning RSVD_EXEC_1 to 512, and
how do I measure the difference?
This would help the system in case of performance issues occurring due to paged
pool fragmentation. When this occurs processes can spend excessive amounts of
time waiting for the Pagedyn and Logical Name Mutexes. Allowing Paged Pools to
have Lookaside lists decreases the fragmentation of its variable freelist and
speeds up allocation and deallocation of commonly used packet sizes.

For example creating and deleting logical names of different lengths and life-spans
can fragment a paged pool.

If you use DCL command "SHOW MEMORY/POOL/FULL" large number(10K) of
"Number of Free Blocks" and "Free Blocks LEQU 64 bytes" indicates that your pool
is fragmented.

Regards,
Pramod.

[Ian Miller.]

SDA can be informative too
SHOW POOL/STATISTICS

[Hoff]

When the documentation here gets updated to correct these omissions (and Jose, it looks like you might be able to directly file an HP problem report around this confusion and around the lack of command documentation?), I'd (also) suggest phrasing this as adding 512 to the value in the RSVD_EXEC_1 system parameter, and not as setting that parameter to 512. Particularly if this parameter is a bitmask, as I might suspect the parameter is implemented. Setting the value to 512 might work now, but if another bit gets lit in the bitmask, things might get weird.