Improving 'catching up'
Posted: Mon May 01, 2017 10:45 am
Hello,
[EDIT] similar posts are describing same problem (see following post)
I joined a public multiplayer game (one of many with same behavior), it took a long time because it was a really big map/save (77MB).
The first step is to download the save : I was downloading at 1.1MB/sec, nobody can do anything because my ISP connexion is the main cause here.
while downloading I was thinking : "damn, It will take soooo much time to catch-up after download"
As expected, the second step of catching up with the server took a long time, but why ?
I tried to figure out :
- At this step, I saw that networking ressources are not used (approximatively a kb/sec)
- I also saw that only one of the six cores of my CPU was used to catch up
- Once connected, I add some freeze (game is still running but I can't interact), when this occurs, I suppose it's because my pc is to slow to catch up with this big map
I absolutely don't know how is your software architecture and how your workflow is going in this step, so my idea is maybe a big big big montain to climb :
Would it be possible to multi-thread the catching-up part ?
[EDIT] similar posts are describing same problem (see following post)
I joined a public multiplayer game (one of many with same behavior), it took a long time because it was a really big map/save (77MB).
The first step is to download the save : I was downloading at 1.1MB/sec, nobody can do anything because my ISP connexion is the main cause here.
while downloading I was thinking : "damn, It will take soooo much time to catch-up after download"
As expected, the second step of catching up with the server took a long time, but why ?
I tried to figure out :
- At this step, I saw that networking ressources are not used (approximatively a kb/sec)
- I also saw that only one of the six cores of my CPU was used to catch up
- Once connected, I add some freeze (game is still running but I can't interact), when this occurs, I suppose it's because my pc is to slow to catch up with this big map
I absolutely don't know how is your software architecture and how your workflow is going in this step, so my idea is maybe a big big big montain to climb :
Would it be possible to multi-thread the catching-up part ?