Feature idea summary

This is the continuation of #7987

We need to have a better idea about what is necessary to move from apps.plugin to ebpf.plugin

The plan is to figure out some more representative mixes of Computation versus I/O, by looking into stress-ng, Transaction Processing Performance Council, fio or vdbench. The aim is also to stress the kprobes that ebpf has implemented.

After having created a couple of scenarios based on the above research, benchmark a system at about 25%, 50%, 75% and 100% load. The system is going to have a 4C/4T CPU so for the 25% load the load is going to be of 2 kinds:

• 25% of multithreaded load
• 25% of single threaded fully busy load (no duty cycle)

No need for the netdata agent to be running. The benchmark is going to affect apps.plugin and ebpf.plugin on their own, to reduce noise.