[closed] as I have new motherboard with at moment only 16 GB RAm = can't make new tests that 'compare'
or the higher the impact, the worse your lag will get!
Read this post how I proofed that it works, and profiler is most easy to use.
Short: I see pipeworks+technic using >50% of my load by profiler, and false them reduce my CPU load by ~50% also, yeah. So I choose lower lag impact mod techage, as options shown in 2nd post.
I have lower lag, or can allow more gamer in now ...
minetest.conf add:
Code: Select all
profiler = true
profiler.load = true
profiler.default_report_format = txt
took a 5.5.0 (later 5.5.1) engine + guess most needed belonging MTG 5.5.0 (5.5.1)
and just one mod ! (if not need dependencies)
and let it run on 'empty' map for 10+ hours
and then as gamer with server privs in the game logged in:
Code: Select all
/profiler save
* make a empty line between two mods
* sort the biggest 10 to top
... then you see what mods make your sever lag !
before:
Code: Select all
instrumentation | min µs | max µs | avg µs | min % | max % | avg %
---------------------------------------------- | --------- | --------- | --------- | ----- | ----- | ------
poshud: | 2 | 35255 | 58 | 0.0 | 97.1 | 3.4
- globalstep[1] ........................... | 2 | 35255 | 58 | 0.0 | 97.1 | 3.4
- /poshud ................................. | 7 | 21 | 13 | 0.1 | 3.2 | 1.0
- on_leaveplayer[1] ....................... | 1 | 85 | 4 | 0.0 | 7.5 | 0.2
flowers: | 0 | 125 | 0 | 0.0 | 3.5 | 0.0
- ABM 'Mushroom spread' ................... | 1 | 125 | 5 | 0.0 | 3.5 | 0.2
techage: | 0 | 8904 | 6 | 0.0 | 79.2 | 0.4
- LBM 'Upgrade reactor' ................... | 14 | 65 | 28 | 0.0 | 1.9 | 0.5
- on_leaveplayer[3] ....................... | 2 | 80 | 5 | 0.0 | 3.0 | 0.2
- /my_expoints ............................ | 8 | 15 | 11 | 0.1 | 0.2 | 0.1
...
Code: Select all
instrumentation | min µs | max µs | avg µs | min % | max % | avg %
---------------------------------------------- | --------- | --------- | --------- | ----- | ----- | ------
technic: | 1 | 50740 | 704 | 0.0 | 99.4 | 37.5
- on_leaveplayer[1] ....................... | 3 | 278 | 10 | 0.0 | 4.0 | 0.4
- on_player_receive_fields[2] ............. | 1 | 41 | 1 | 0.0 | 2.8 | 0.0
- ABM[1] .................................. | 45 | 166 | 71 | 0.1 | 2.7 | 1.8
- ...
pipeworks: | 0 | 95305 | 804 | 0.0 | 99.1 | 21.5
- ABM[4] .................................. | 8 | 3458 | 425 | 0.1 | 39.8 | 9.5
- globalstep[2] ........................... | 1 | 36004 | 3 | 0.0 | 97.5 | 0.2
- globalstep[1] ........................... | 1 | 95294 | 807 | 0.0 | 99.1 | 21.4
- ...
wielded_light: | 0 | 1717217 | 219 | 0.0 | 97.9 | 6.7
- globalstep[1] ........................... | 1 | 48164 | 85 | 0.0 | 97.9 | 3.3
- on_punch[1] ............................. | 22 | 274 | 52 | 0.0 | 13.3 | 1.9
- on_activate[1] .......................... | 10 | 17075 | 166 | 0.0 | 35.7 | 4.3
- ...
...
And then you can decide to keep and know, or to change to lower impact mods = less lag / more gamer.
example of CPU usage with only empty world, no gamer, no mods on different ones:
as similar to more load, also CPU getting more busy and reach 100% is no fun gaming ...
* 15.9 - 30.9 % - I7-4770 3.4 GHz - 4 Core/8 Thread - 32 GB Ram (minetestserver 200 mods, 20 gamer)
* 4.0 - 5.3 % - I7-4770 3.4 GHz - 4 Core/8 Thread - 32 GB Ram (minetestserver, no mods, no gamer)
* 3.3 - 5.3 % - I7-4770 3.4 GHz - 4 Core/8 Thread - 32 GB Ram (minetest --server, no mods, no gamer)
* 0.7 - 1.5 % - Intel E1, 1.4 GHz - 1 Core / 2 Thread - 4 GB-Ram (minetest --server,, no mods, no gamer :interesting)
then take a /profiler save extract.
So you admins, will have a imagination about what happen when you add one of this measured mods to your server and lag.
I think, I should get the most activate mods, even their items are not used, for a first impression where to be careful, choosing an mod for a server ?
but also never forget to check out tuning in:
* hardware (most more RAM (slots better 4*4GB RAM than 1*16) and faster disc as NV, SSD than HDD)
* OS tuning, like enable/better cache usage, no swappiness
* database better/faster postgresql (with set more Cache) -- than maybe sqlite3
* minetest engine tuning, as keep data longer, less active mapblocks, lower globalstep ?
* mod tuning (take the one less impact, reduce manual as ABM interval/chance)
as a lot of hints and practical changes all over my server Thread : Asia Thailand and on tuning.minetest.one
btw, this idea is so old ... and never done yet: