Optimization for Large Servers aka PMMP vs. Genisys and The Future

If you want to handle > 80 players comfortably I think SSD and a dedicated server, not a VPS, are essential

 

Ikr!

 

Ugh, Pocketmine-Soft was not that good. I always had it when I was using Pocketmine and most bugs weren't actaully fixed. PMMP is like PocketMine dev and soft mixed into one. Although, maybe a non-programmer friendly/GUI version might work. Like how they have it on the PC servers.

 

PocketMine-Soft wasn't supposed to fix any bugs. And how will a GUI help..?

 

I'm pretty sure that Usain Bolt won't be ever possible to break his own world record by significant difference without upgrading equipment or using high tech exoskeleton.

 

For people who have absolutely NO idea how to use CMD/BATCH and are just trying to setup a server for their friends because realms is too costly.

 

Over all: what about making benchmarks and REAL comparasions between 'spoons'/forks and PMMP?

1. Player tests: 0, 30, 60, 100
2. Block placing tests
3. Network speed tests (compression etc)
4. Calculation tests
5. Plugin speed tests
6. Tests run on different systems / setups

This would provide the truth about the power.

 

I thought about e.g. Manyworlds, WorldEdit/MagicWE, they should be kind of the same speed on any core

Simply $i++. (actually something more resource intense.) Maybe some sqrt() calculations?

Approved point.

 

I thought we were talking about software for high-performance, not easy use.

 

Oh yeah, sorry for going off-topic

 

Possible; but not an easy task.

"REAL" automated performance testing is something that big game companies (e.g. EA) have spent millions of dollars building code for. Even their code tends to fail. In other words; the performance against the artificial test code looks good, but when the game launches to real players there is a big surprise. We have seen this with many major multiplayer game launches over the last 10 years (e.g. Battlefield 3/4). It is a very difficult thing to do well. In the end it's usually best to just test live with real humans; and tweak the code from there.

Specific to a large Factions server, what would need to be simulated would be 60 players including:
- A certain percentage of players doing PvP with a wide variety of weapons (mix of projectile as well as enchanted melee).
- A certain percentage of players doing lots of building (block placing).
- A certain percentage of players doing heavy destruction and collection of blocks (mining).
- A certain percentage of players moving erratically in a central shop/town area. Tapping signs. Showing off their skins.
- A certain amount of mixed crop growth within areas of the world that contain active players. This can be substantial for a 20GB world.
- A certain amount of water/lava flow and weather effects (snow gathering) within areas of the world that contain active players.
- A certain amount of redstone machines of various complexity running within areas of the world that contain active players.

Of course adding key plugins to the equation would be vital. The simulation should at least include a plugin that check each block place/destroy against some sort of "land ownership" plugin; as this is common in most servers (e.g. EconomyLand).

I probably have missed many things. I'm sure not a small number of CS PHD candidates write their thesis papers on these types of simulation problems. They are non trivial.

 

I cried for joy when I found that out. I was so confused when my lake house lagged and I had to abandon it.