0% read hit rate on XFC cache for RMS indexed file being read sequentially using C RTL fgets()?

[P Muralidhar Kini]

Hi Mark,

>> I've been informed that the volume does get backed up, and it is backed
>> up using Legato.
This is interesting. VMS backup I know of sure takes care of skipping the
XFC cache and hence avoids thrashing of XFC cache.
What we need to find out is whether the Legato backup application also do
something similar to VMS backup and skip the XFC cache.
If not then the Legatobackup will trash the XFC cache and wipe out the data
already present in the XFC cache.

I was thinking of some way by which we can make the Legato backup to skip
the XFC cache. In the existing versions of VMS there is no way to
dynamically enable/disable XFC cache for mounted volumes.
This feature would however be provided by the upcoming VMS V8.4 Release.

For the time being, If its feasible, you can try this test -
Dont backup the disk using legato backup for one day and monitor the
performance of the application the next day. Is it XFC statistics the same
when the application starts or is there any difference.


>> However, it does show that the cache drops in size to ~400MB in places.
Yes correct. we can clearly see the XFC allocated memory shrink number of
times. It does get as low at ~400MB. This is a indication that system has
got memory starved and has requested XFC to shrink and release some memory.


Some statistics from "SDA>XFC SHOW MEM" output -

>> MMG callback count : 1216058
This is the number of times System has requested XFC to trim down its memory.

>> Expand attempts refused : 268666
This is the number of times XFC did not get memory from the system when it
tried to expand its memory.

All these data points to the fact that the system gets memory starved at
times due to which it requests XFC to trim down. This causes XFC to depose
contents of the cache in order to free up memory.
It would be interesting to see your analysis on the memory usage of the system.

Add to the legato backup application which can also potentially thrash the XFC
cache unless it specifically takes care to skip the XFC cache (like VMS backup
for example).
Does the Legato backup application issue IO's such that it skips the XFC cache?

I will look in to the "SDA>XFC SHOW HIST" output in more detail for more clues.

Regards,
Murali