Felipe Ortega schrieb:
I completely agree with Tim's answer. Proper hard disk resources (in terms of speed) are critical.
MySQL INSERTS don't overload the whole process very much. I usually do the final inserts to MySQL separately and it doesn't take more than the 5-10% of the whole processing time in any case.
I believe that Perl parser is somewhat faster than mwdumper (and definitely, faster than the research version of my WikiXRay new parser; I still have to check them against the next standard version of my parser).
However, you should take into account that, depending on the language you're processing, certain revisions could have a *very big size*, and inevitably any parser, no matter how improved or multithreaded it is, will spend a considerable time to process them (I mean, when we look at the aggregate number of tasks the parser must face off).
It usually takes a week or so for me to load back to MySQL the whole dump of the English version, and that's with a big server with 2 Opteron 2GHz (double-core each one), a lot of fast memory and a RAID 6 array of 8 fast SATA-II disks.
MySQL configuration will be critical later, when you try to "play with your data". I recommend you www.mysqlperformanceblog.com for that. You'll find very useful hints there.
Thanks a lot Felipe, seems as if my two weeks importing time are OK so far.