Talk:Vhosting possibilities
Jump to navigation
Jump to search
- Why tile the ocean into sim-sized regions at all?
- For shoreline cruises or events, just extend the existing shoreline sims further into the water. They don't need to be square.
- And as regards the high seas, the workload is likely to be so small that the entire ocean could be handled by one sim initially, then sliced into two as demand grows, etc.
- Alternatively, and very preferably .... take this opportunity to virtualize zones at long last, at least for the ocean. Instead of partitioning ocean acreage across static resource tiles, distribute the workload generated within that total sea area across a dynamic pool of region processors. That way all the CPUs are always contributing to grid performance in event hotspots, instead of most of them being largely idle in residential sims. (In other words, the way that SL should have been designed in the first place if it had been designed for full scalability including events.)
- Tiling the ocean dynamically really qualifies as compounding a bad thing further. Fully static allocation did at least offer simplicity, if not scalability. This does neither. --Morgaine Dinova 20:24, 24 September 2007 (PDT)
- Virtualization is a powerful concept, and extremely useful as a sysadmin tool, particularly for migration. However, it doesn't help in the ongoing scalability exercise much (at least not for event scaling), because it provides downscaling (ie. running many sims on one box) instead of upscaling (spreading the activity of one hot event sim across many boxes to carry the load). --Morgaine Dinova 20:36, 24 September 2007 (PDT)
- There is now a proper Brainstorming section on this topic, Virtualization of regions. --Morgaine Dinova 04:23, 6 October 2007 (PDT)