Game Development Reference
In-Depth Information
the priority queue and can result in a much more fluid and speedy load in the
subsequent play tests.
2.4 Bluer Skies
The system described above has been battle tested with projects that each of us have
worked on. Let us now entertain how the whole process can be improved. Going
back to the first section, the simple calculus introduced can actually be integrated
into the tool chain. There can be two different modes of loading assets in the
game, one which loads all assets at startup and one that simulates streaming loads.
The first, which loads all assets, can be used for data collection during game play-
throughs. A table is kept of what assets are requested and at what time they are
requested. Furthermore, each table has metadata to describe the player location,
level being played, and other characteristic features that allow these reports to be
grouped appropriately. Group data can then be analyzed for patterns and graphed
for review by the development team. Based on a target bandwidth, the data can
help pinpoint problem areas using our
and Ω metrics. A game designer can adjust
the length of certain gameplay situations to offset when other assets are introduced.
Alternatively, proxy assets or a lower level-of-detail for assets can be used in place
of assets that take longer to load. The group data can be averaged to generate an
optimal load scheme that answers both which assets should be bundled together
and when a bundle should be loaded. Asset bundles no longer need to be explicitly
created, and, instead, a manifest is generated that maps assets to the asset bundles
that contain them. Gameplay engineers are relieved from having to know which
bundles to request to load a specific asset. Finally, these optimal load schemes are
fed back into the game, allowing for a proper load scheme to be picked based on
the current characteristics of the play session. All of these improvements mean that
over time the game can approach zero load times in a streamlined fashion.
ω
2.5 Conclusion
Digital distribution is still a fairly recent development in the game industry, and
although it has gained a great deal of momentum, its true potential has yet to be
realized. Preloading a portion of a game to allow a player to start playing early
is a good start; however, instant play should be the ultimate goal. To advance
progress towards this goal and have online game load times approach zero, asset
streaming must become a chief game design concern and one that is included at
the very beginning of production. Tools for visualizing the asset download process
and managing grouping and order of assets and possibly a whole integrated tool
chain that refines this process over time is necessary to allow developers to create
the great online games of the future.
