i facing this problem 3 weeks ago until now i’m running 4 core ubuntu 16.04 16gb ram and i followed the performance tuning wiki page notes but still facing performance issues with 11 pos machine live plus 5 users at backend
and here is the htop output
Hi
Lets talk English 
What Performance issues are you facing ?
This is very similar to a set up for one of my companies
How many items?
How many customers?
2000 item
1450 customer
40000 invoice
11 live pos machine
5 backend users
Have you analysed PROCESSLIST?
Try tuning your innodb_buffer_pool_size. Try first with 2GB while innodb_buffer_pool_instances is set to 2. See if there is a change in performance or not
its set to 12G but innodb instance are not found at settings.cnf at /etc/mysql/conf.d
you can see here that gunicorn also eating the cpu
No no 12G is way too much unless ERPNext and MariaDB are running on different servers altogether.
the recommendations at performance wiki page said 70-80% from total ram and i have 16g ram
No that would be a good setting for those who are running seperate servers for ERPNext and MariaDB i.e. an ERPNext instance with a remote database setup. Try tuning it down to something lesser according to your database size.
However innodb_buffer_pool_size won’t solve the high CPU usage by gunicorn. I’ll see if this has been raised on the forum and get back to you.
EDIT : I found this link which has similar POS problem like yours
It is recommended by a user to try and experiment with varying number of gunicorn workers. You can find out the gunicorn worker numbers in your currentsite’s site_config.json
3000 items and
2500 customers.
