Need to see output of free -m taken when memory use starts to staircase up as shown in your graph. Possible it is just memory allocation for buffering/caching (which is expected), but the fact that you start to experience problems would seem to suggest otherwise.
While memory is staircasing up, first step is you need to try to determine if the memory utilization is in process space or kernel space. Easiest way to do this is run top and hit capital "M" which will sort processes by top memory utilization. Do you see any processes sucking up more and more memory? If so check the corresponding elg file for that process to gain insight into what is going on, for example if the daemon name is fwd, check the $FWDIR/log/fwd.elg log file (or possibly $CPDIR/log/fwd.elg depending on which daemon it is).
If you don't see any processes obviously hogging memory in top, it means that the issue is in the kernel, and the next step would be to run fw ctl pstat to try to figure out what specific type of memory is being excessively consumed (i.e. hmem, smem, kmem).
Obviously what to do about whatever you find is another matter entirely, but hopefully this has given you a few places to look and start narrowing down where you should focus your efforts. Loading the latest GA Jumbo HFA is also a very good idea, as these will frequently include fixes for memory leaks.
Gateway Performance Optimization R81.20 Course
now available at maxpowerfirewalls.com