Memory usage in NGx R65

[cciesec2006]

I have two different NGx R65 clusters running SPLAT on IBM x3650 servers,
dual quad-core processors 3.16 Ghz with 4GB RAM. All of the servers
are running NGx R65 with HFA_02 and hf_249. These gateways are
managed by SPLAT Provider-1 NGx R65 with HFA_02 and hf-249.

on cluster#1, there is NO traffics running through these gateways,
when running top, I see only about 200Mb of FREE memory on both
gw1 and gw2.

on cluster#2, there is NO traffics running through these gateways,
when running top, I see about 3.8GB of FREE memory on both gw1 and
gw2.

Question is, why such big differences between these two clusters in
terms of memory usage?

When I applied HFA_30 on both cluster #1 and cluster #2, I see only about
100MB of FREE memory on ALL of the gateways but that is another
discussion for another day.

Can anyone explain how NGx R65 allocates memory? It makes no sense to
me. Thanks

[Thorpuse]

The question is probably not how it allocates memory, but how it displays free memory. As someone from CP tried to explain, VPN-1 pre-caches memory and marks it as used for certain functions, even though it's actually free. I'd like to find the right set of commands to show actual memory use, because it's standard mode of operation looks like it's sucking memory like no-one's business.

I'd suggest a support call about that though - sounds like something is going funny at the OS level.