perfSONAR User Q&A and Other Discussion

Hey.
Server: VMware 12 cpu 64GB ram.
My server should work as a reflector for ~ 21000 twamp sessions (1 UDP packet
per second)
I have the following problem, when the number of twamp processes reaches
~62000

[root@vlg-gdc-tst02 ~]# ps --no-headers auxwwwm | cut -f1 -d' ' | sort | uniq
-c | sort -n
      2 ntp
      2 owamp
...
    503 root
  59190 twamp

the server stops executing commands and responds: "-bash: fork: Cannot
allocate memory"
In the logs I see "twampd [76220]: FILE = owampd.c, LINE = 572, fork ():
Cannot allocate memory"
At the same time, there is still a lot of free memory:

[root@vlg-gdc-tst02 ~]# free -g
              total        used        free      shared  buff/cache
available
Mem:             62          21          38           0           3
39
Swap:             7           0           7

Question: is it possible to configure the server for such a load or only
increase CPU?

I already have 2 HP bare-metal servers running 24 cores 32GB of RAM that can
handle the load of 18K sessions.