Wheelerof4te wrote:^Average non-gamer users won't notice much difference in performance.
That's a good point. I just wanted to see first hand the difference with pre- and post-patch performance. You are right that the actual impact to average user is minimal. Gamers also shouldn't be too much affected, unless the game is especially CPU intensive. Like you said, most impact will be suffered by business users, service providers and users who need to squeeze out every bit of performance from their hardware.
So I ran following sysbench tests, which seemed to show some change. I ran three tests before and after the patch, below results are my own redacted summary of all six tests. System load and processes for all tests was the same, but I doubt my test setup would fulfill scientific standards..
Thought I'd share them with you anyway, maybe someone's interested.
EDIT: Disclaimer, I'm not really big on system benchmarking. There probably is lot better ways to test it. Maybe someone could comment on this?
Code: Select all
Hardware:
Intel i5-3570k @ 4.4Ghz, 8Gb DDR 3
Tests run with:
$sysbench --batch --batch-delay=5 --num-threads=2 --max-requests=50000 --test=threads run
Pre-patch, with kernel 4.9.65-3+deb9u1
Pre-patch Test 1-3
total time: 5.3247s - 5.4139s
total time taken by event execution: 10.6455 - 10.8239
per-request statistics:
min: 0.18ms - 0.19ms
avg: 0.21ms - 0.22ms
max: 0.57ms - 0.92ms
approx. 95 percentile: 0.23ms - 0.24ms
Threads fairness:
events (avg/stddev): 25000.0000/2.00
execution time (avg/stddev): 5.3228 - 5.4119/0.00
Post-patch, with kernel 4.9.65-3+deb9u2
Post-patch Test 1-3
total time: 8.7143s - 8.8911s
total time taken by event execution: 17.4244 - 17.7782
per-request statistics:
min: 0.32ms
avg: 0.35ms - 0.36ms
max: 0.97ms - 1.03ms
approx. 95 percentile: 0.36ms - 0.37ms
Threads fairness:
events (avg/stddev): 25000.0000/2.00
execution time (avg/stddev): 8.7122 - 8.8891/0.00