This article explains why the output from show system process extensive reveals memory utilization is increased by 10x after upgrading to BSD 15.1.
show system process extensive
Users who monitor daemon memory utilization on legacy (non-BSD) Junos using the value “ Size ” under show system process extensive should now monitor “ RES ” on BSD-based Junos 15.1x releases and later.
In Junos, every daemon that matches the configuration allocates a block of 700+ MB RAM associated to them. In legacy (non-BSD) Junos, it only projected a subset of the actual memory usage that is potentially used by a process in the output of show system process extensive under the " Size " column. It also hid the configuration database and other internal information that is mapped to every daemon. However, in BSD-based Junos 15.1, it does not project the subset anymore. It shows what is polled from the FreeBSD directly without hiding any details like it did earlier. You may see a large amount of inactive memory; this is memory that is cached (ready to be reused). This is not a problem.
If you are are monitoring daemon memory utilization for potential leak or abnormal utilization, you should look at RSS (Resident Set Size) under show system process detail output. This shows the actual memory (RAM) used by the process.
show system process detail
labroot@MX960-r2010-re0> show system processes extensive | grep "PID|mib2d" †last pid: 16521; load averages: 0.13, 0.18, 0.16 up 0+06:06:42 18:22:04 PID USERNAME THR PRI NICE SIZE RES STATE C TIME WCPU COMMAND 14615 root 1 20 0 776M 19164K select 0 0:20 0.59% mib2d <<<<<< Notice 776M, in pre-BSDx it would be approximately around ~70M 14069 root 1 20 0 719M 5544K select 2 0:00 0.00% appidd labroot@MX960-r2010-re0> show system processes detail | grep "14615|RSS" † PID UID PPID CPU PRI NI RSS WCHAN STARTED TT STAT TIME COMMAND 14615 0 14031 0 20 0 19164 select 12:26PM - S 0:19.83 /usr/s
2018-09-27: Minor, non-technical edit