A post I made at UO bears repeating here:
========
What I’ve heard of the server software (from UU’s perspective) would not require too much major rework for scalability I think. You’re connecting to only one ‘Age’ instance at any one time anyway. So…
Put each Cavern instance (I heard 5? More likely 2-3 for ‘Production’) on a separate server or CPU/core; That should allow better connectivity and up to 100 or so instead of the average 45 (though total in shard was almost over 100 once?). Terrapin’s fog concept works well in this respect too; Never quite understood why you’d send/rcv movement data for somebody you can’t even see.
Relto/Nexus and anyone playing the Ages in ‘Prime’ fashion moved to one scalable cluster; Instances spawned should balance similar to game instances on something like Diablo II Realm Games. And ‘hoods on another scalable cluster. Since KI comms is mostly text, a meta-process similar to an IRC-daemon can be stuck on a medium-grade box.
I’d be curious to know what is getting reworked on Plasma. Because you could still do all this on *one* server and be compatible to the existing format. The all important crux is how much the network communication code got ‘fixed.’
========
I think I’ve spent too much time in SL, and as a Unix SysAdmin, to come up with something like this…
;)
:link: