As you’ve asked for real time information, such as process count per core, I’m assuming you’re asking for monitoring tools rather than benchmarking. Erlang comes with some built in tools, there’s also some you can add through hex.

etop

Command: :etop.start

Starts an in terminal UI similar to the unix top command. Does for processes in the VM what top does for processes in eg a linux system.

observer

Command: :observer.start

Starts a windowed application that displays a huge amount of information on the running VM, including memory, processes, load, process trees, and lots of other stuff.

observer-cli

An upgraded version of etop with lots of information, can be installed through hex. Possible to run remotely.

recon

This is a collection of tools used to get more information on your system, useful for debugging problematic systems.

NewRelic does have an Elixir agent…

So yes benchee is more for measuring performance before something goes live or for testing out multiple approaches locally, from what I understand you want to have more of an introspection into the running system. @jola offered multiple great solutions there.

Adopting Elixir, which is great, has a chapter about Metrics and instrumenting your system.

Other projects they mention as being used in the wild:

• exometer for pushing performance data to services such as Data Dog
• football addicts has their own write up of how they use Telegraf, InfluxDB and Grafana
• Prometheus and Grafana is yet another choice

I like whole Application Performance Monitoring Solutions in the SaaS space such as:

• AppSignal
• Scout
• wombatoam
• there are some community developed agents for NewRelic, not sure how good they are 1 2

And for even more options you could take a look at some of our previous threads on the subject:

• How do you monitor / profile your live Elixir / Phoenix apps?
• What do you use to monitor your Elixir app?

There also is OpenCensus for Erlang that has few integrations already and is provider agnostic. Telemetry.Metrics provides OC-inspired set of metrics as well.

And please, could we stop promoting that abomination of library called exometer? It sends shivers down my spine on just mention. Not only it was hellhole to compile it with Mix it also is highly outdated (really, no metric tags in 2019?). There are much better and simpler solutions already available (for example vmstats and statix for the example). And with raise of the telemetry tracking metrics will be even easier.

I’ve never used exometer so I couldn’t tell, but calling a library an “abomination” isn’t the right way to go or talk about open source libraries. People took care and time to build it offering it to you for free - nobody forces you to use it. You could even improve it. You have your reasons for not using it, that’s fine but you don’t have to resort to insults. That’s what contributes to burnout for maintainers & others. Also that’s really not something I want to see as a tone in the community as it also keeps people away from open source, as they are often afraid of that exact thing happening to them.

Never mind that the library is mentioned positively in a book (again Adopting Elixir) written by people respected by lots of us which was published about a year ago and where one of the authors is a lead developer at a big elixir company (bleacher report) using exometer apparently happily (otherwise I doubt he’d have included it in the book). Might not work for you, but might work for others.