[MUD-Dev] Is database access a bottleneck? Prevayler.

[Kwon Ekstrom]

From: <ra-whiteside at attbi.com>

> This is a replacement to normal database technology for managing
> data in a persistent, reliable, crash-recoverable manner.  The
> proponents of this seem to have benchmarks showing it to be a few
> thousand times faster than MySQL, for example. (DAOC uses MySQL?)

> At any rate, the underlying Prevayler idea is applicable in any
> language, it's just particularly easy to implement in Java because
> of the object serialization available in that language.  

> Well, that's about it.  Is this interesting at all?

I've checked out the site and looked into it, and while I find it to
be a rather interesting piece of software, I don't think it's an
adequate replacement for an Database.

It's a good idea from the business viewpoint.  Business objects tend
to be pretty simple, and are mostly for keeping track of
transactions within the business.  Whereas any decent RPG probably
has dozens (if not hundreds) of different data structures, a high
percentage of which are pretty complex.  Data recovery and
Synchronization are vital to business structures, while most RPGs
can get away with losing some data during a crash (although the
players might complain about losing some of their experience and/or
equipment...)

I don't know about any commercial MMORPG's but I can tell you that a
home brewed MU* changes constantly (a good one anyway) and Java
Serialization, while good at keeping an accurate image of an object
is terrible for this sort of thing... Any change in a java object
means you can't load previous versions of the same object (really
bad news if you want to keep your data).

Also it seems to require that you generate Command objects.  Which
for a small system is a good idea, but as the complexity goes up
you'd need to create more and more command objects, and often used
command objects would need some sort of pooling mechanism to prevent
object churn.

I just glanced at it, so there may be some solutions to these
problems, but I doubt it's entirely feasible for a major server.

-- Kwon J. Ekstrom


_______________________________________________
MUD-Dev mailing list
MUD-Dev at kanga.nu
https://www.kanga.nu/lists/listinfo/mud-dev