Cross post: https://forums.frontier.co.uk/showt...Event-Server-quot-model?p=5467231#post5467231
The Salome event made it pretty clear (again) that the current P2P model does not work well with large numbers of players in the same place all trying to do the same things. Most likely this is down to the vagaries of network traffic and the high variance in internet connection speeds etc. I suspect, therefore, that there is literally nothing that can be done at this time to "fix" the issue, except to wait until fibre internet is ubiquitous.
However, in the meantime, I wonder if an "On-demand event server" model might be worked into the current model and whether this might improve things for large scale events. I am imagining this event server would spin up on AWS as the other servers do on demand. They could spin up as soon as more than N players were present in a given system and could move between system as required (much as a player does). Given pre-planned events some could spin up early in those systems to handle the expected demand.
These event servers would fit into the existing model by behaving exactly like a player P2P endpoint does except that the instancing algorithm would assign them the highest possible priority for being the instance controller. The end result would be that they would essentially always be in control of instances in busy systems, instead of a player endpoint (with its associated connectivity/delay/etc issues).
As a developer myself I know that this is likely to be far more complicated than I can possibly imagine, but my motivation here was to suggest something which might be possible to engineer within the current P2P model (which I suspect will never change as it's far too ingrained into everything), which might not be an astronomical amount of work, and which might improve the stability/consistency of the player experience, especially for high profile events. I don't know how much "work" the P2P endpoint has to do, and whether it's even this endpoint or the other server transactions which is causing the bad experience.
p.s. I am a server side developer in the UK (not far from Cambridge) with nearly 20 years experience (tho none of it in the games industry).. just sayin
The Salome event made it pretty clear (again) that the current P2P model does not work well with large numbers of players in the same place all trying to do the same things. Most likely this is down to the vagaries of network traffic and the high variance in internet connection speeds etc. I suspect, therefore, that there is literally nothing that can be done at this time to "fix" the issue, except to wait until fibre internet is ubiquitous.
However, in the meantime, I wonder if an "On-demand event server" model might be worked into the current model and whether this might improve things for large scale events. I am imagining this event server would spin up on AWS as the other servers do on demand. They could spin up as soon as more than N players were present in a given system and could move between system as required (much as a player does). Given pre-planned events some could spin up early in those systems to handle the expected demand.
These event servers would fit into the existing model by behaving exactly like a player P2P endpoint does except that the instancing algorithm would assign them the highest possible priority for being the instance controller. The end result would be that they would essentially always be in control of instances in busy systems, instead of a player endpoint (with its associated connectivity/delay/etc issues).
As a developer myself I know that this is likely to be far more complicated than I can possibly imagine, but my motivation here was to suggest something which might be possible to engineer within the current P2P model (which I suspect will never change as it's far too ingrained into everything), which might not be an astronomical amount of work, and which might improve the stability/consistency of the player experience, especially for high profile events. I don't know how much "work" the P2P endpoint has to do, and whether it's even this endpoint or the other server transactions which is causing the bad experience.
p.s. I am a server side developer in the UK (not far from Cambridge) with nearly 20 years experience (tho none of it in the games industry).. just sayin