R
rstanley
Guest
In responding to a posting this morning by @skbansal2221 "which is faster?", inquiring if the GUI or CLI is faster, it got me thinking about a related question.
How does the overhead of a running, but relatively idle, GUI, in my case KDE, affect the speed of the CLI? Out of curiosity, I decided to run a VERY simple benchmark on my older desktop system. Most Data-Canter Linux servers are running either with the GUI turned off, or never installed at all. Is there much of an improvement if a GUI is not running?
My System specs:
Dell Vostro 200, Pentium Dual Core, 2 GB memory.
Debian Jessie, updated as of this morning.
Linux ????? 3.16-3-686-pae #1 SMP Debian 3.16.5-1 (2014-10-10) i686 GNU/Linux
I installed sysbench through the Debian package manager:
"sysbench 0.4.12: multi-threaded system evaluation benchmark"
My KDE (Runlevel 2) had three apps running idle, Iceweasel, Dolphin, and emacs. (Obviously, active applications running in the GUI, such as a running video player, would affect the results.)
Command line in both tests:
Results with KDE active, with the command run in the "F1" console, NOT from within a KDE terminal:
Results run from the console, in Runlevel 1, with no GUI active:
This is only one execution of each test, on one system. Other runs of both, will produce slightly different run times.
Different computers from different manufacturers, with different processors, different Distros, different kernels, different amounts of memory, etc... would affect the results.
Comparing the two "Total Times", a difference of 0.559 seconds, over almost 56 seconds, is far more insignificant than I expected!!! Bottom line, in the future, I may not bother to turn off the GUI on my servers! You gotta love Linux! ;^)
How does the overhead of a running, but relatively idle, GUI, in my case KDE, affect the speed of the CLI? Out of curiosity, I decided to run a VERY simple benchmark on my older desktop system. Most Data-Canter Linux servers are running either with the GUI turned off, or never installed at all. Is there much of an improvement if a GUI is not running?
My System specs:
Dell Vostro 200, Pentium Dual Core, 2 GB memory.
Debian Jessie, updated as of this morning.
Linux ????? 3.16-3-686-pae #1 SMP Debian 3.16.5-1 (2014-10-10) i686 GNU/Linux
I installed sysbench through the Debian package manager:
"sysbench 0.4.12: multi-threaded system evaluation benchmark"
My KDE (Runlevel 2) had three apps running idle, Iceweasel, Dolphin, and emacs. (Obviously, active applications running in the GUI, such as a running video player, would affect the results.)
Command line in both tests:
Code:
$ sysbench --test=cpu --cpu-max-prime=20000 runResults with KDE active, with the command run in the "F1" console, NOT from within a KDE terminal:
Code:
Running the test with following options:
Number of threads: 1
Doing CPU performance benchmark
Threads started!
Done.
Maximum prime number checked in CPU test: 20000
Test execution summary:
total time: 55.9829s
total number of events: 10000
total time taken by event execution: 55.9784
per-request statistics:
min: 5.59ms
avg: 5.60ms
max: 8.42ms
approx. 95 percentile: 5.63ms
Threads fairness:
events (avg/stddev): 10000.0000/0.00
execution time (avg/stddev): 55.9784/0.00
Code:
Running the test with following options:
Number of threads: 1
Doing CPU performance benchmark
Threads started!
Done.
Maximum prime number checked in CPU test: 20000
Test execution summary:
total time: 55.9225s
total number of events: 10000
total time taken by event execution: 55.9183
per-request statistics:
min: 5.59ms
avg: 5.59ms
max: 7.92ms
approx. 95 percentile: 5.60ms
Threads fairness:
events (avg/stddev): 10000.0000/0.00
execution time (avg/stddev): 55.9183/0.00This is only one execution of each test, on one system. Other runs of both, will produce slightly different run times.
Different computers from different manufacturers, with different processors, different Distros, different kernels, different amounts of memory, etc... would affect the results.
Comparing the two "Total Times", a difference of 0.559 seconds, over almost 56 seconds, is far more insignificant than I expected!!! Bottom line, in the future, I may not bother to turn off the GUI on my servers! You gotta love Linux! ;^)
