Welcome to the Linux Foundation Forum!

Difficulties of system tuning

the_bro
the_bro Posts: 20
in LFS201 Class Forum - Discontinued

The lesson alludes to the fact that memory tuning is complex at best and based on the overwhelming number of available parameters, I'd like to suggest notwithstanding malfunctioning hardware/software, that load balancing should be used as a short-term tool to solve performance issues.
In other words, restrict application usage until more/better hardware/software can be thrown at the problem(s).
Trying to manually tune today's operating systems is akin to a pilot thinking she can better manage an A380 jumbo than the automated systems included in the aircraft.

Perhaps time can be better spent trying to solve the Andries Brouwer conundrum :) !

Thoughts, anyone?

Comments

  • coop
    coop Posts: 917

    Some manual tuning will always be necessary. Operating systems, compilers, libraries etc are all written by human beings and require being continually revised and tuned. I can remember cases where no matter how hard I tried to customize and tune an application I could not wreak any more out of it than the system and compiler could. But I can also remember cases where I had to override those very same factors to increase performance, throughput, use less energy etc. There is always a competition. Plus the OS is always designed for a more generic workload than a specific system or application may need.

  • cfuchs
    cfuchs Posts: 15
    edited January 2022

    Reminds me of comparing C++ to Java 😉 But consider that Linux runs in the cloud as well as in embedded systems (on virtual hardware as well as on bare metal); I think your point is fully valid for virtual systems; on restricted systems, some system engineer is probably happy to have these tuning opportunities.

Categories

Upcoming Training