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

Daniel.Harman at barclayscapital.com Daniel.Harman at barclayscapital.com
Thu Dec 12 11:49:48 CET 2002


>From ra-whiteside at attbi.com

> I recently came across an idea called "Prevayler."
 
>    http://www.prevayler.org
 
> 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?)

Yuck!

Its just an in memory database with logging, so its not exactly a
new idea.  I only scanned the faq, but it has several enormous
problems that I can see:

  - No ad-hoc query support. Having to write code structure
  navigation code to do any data-mining is an awful burden.

  - Its not clear how you would share the access across multiple
  clients, and how it would scale should you try to.

  - It doesn't support roll backs. This appears to make the
  transactions awfully limited, but I may have failed to grok his
  explanation of why transactions aren't needed.

  - Its completely dependant on using Java.

  - It doesn't seem to have intrinsic support for constraints, so
  ensuring data integrity is entirely down to your code. That's not
  a good idea imho.

Some very smart people have spent a very long time writing databases
that do some very smart stuff. I don't find it credible that a few
guys with a logging in-memory database are going to create something
that does the job better, and from what I've had time to read, its
very feature light. From my perspective, the lack of ad-hoc query
language kills it dead before even getting to the other
points. Telling me I can write my own code to search the object
store doesn't cut much ice, and could be incredibly tricky if your
search doesn't match the structure of the data well (which would
admitedly make it a slow query in SQL, but at least it would be
possible).

As to the comparison with MySQL, whilst it's a reasonable tool as
far as it goes, I don't really consider it a heavy weight database.

I'm still more interested in the apparently free to use Microsoft
database thingy that's based on SQLServer code (bleh, the name
escapes me at the moment), but lacking its graphical front end. I've
not had a chance to have a proper poke and see whats missing (I know
replication etc. are out though). If its really free to use though,
it would make a fantastic foundation for a MUD.

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



More information about the mud-dev-archive mailing list