>> (1) The metrics we want to take: cpu, ram, net, disk. >Definitely. Can we elaborate a bit on what sort of metrics we'll intend to gather for each system module?
Todd, thanks for response. We wish to get usual stats: - cpu - its utilization in % - ram - used and total - net - sent/received bytes/packets - disk - I/O operations, used/total space >> (1) The metrics we want to take: cpu, ram, net, disk. > I think having these 4 is enough. However, how we are going to gather these > metrics (i.e. overall instant CPU usage at sampling time vs. average CPU > usage of runner process between each sampling; total disk usage vs. > disk IO etc.) will probably depend on user needs. So maybe for starters, > we choose one method for each, but design flexibly so that we may change > the methods we use in the future, or even might let the user decide. These are good points, Yigit! I agree with that for the beginning we can implement simple solutions. Later we will do further improvements. Regards, Askat On Fri, May 30, 2014 at 4:58 AM, Yigit Sapli <[email protected]> wrote: > 2014-05-29 11:41 GMT+03:00 Askhat Asanaliev <[email protected]>: > > > > Alex, Yigit, > > > > I would like to clarify the main > > points before starting the implementation. Just to be sure that there is > no > > confusion or I am not missing any points. > > I think this is pretty solid. > > > (1) The metrics we want to take: cpu, ram, net, disk. > > I think having these 4 is enough. However, how we are going to gather these > metrics (i.e. overall instant CPU usage at sampling time vs. average CPU > usage of runner process between each sampling; total disk usage vs. > disk IO etc.) will probably depend on user needs. So maybe for starters, > we choose one method for each, but design flexibly so that we may change > the methods we use in the future, or even might let the user decide. >
