>[O'Toole, Eamonn] This definitely sounds like it could be an issue. There is >just one container_stat entry >confirmed by sqlite3_analyzer output which I'll >post later). So you're saying that the single container_stat table :entry is >potentially being relocated very frequently the closer the update is to the >beginning of the db?
Yes....you didn't say what your data flow is...but since it's account-based I assume you have a bunch of accounts that get preloaded. So the first n-thousand records are Size1. You then start updating each of those...none of them are big enough...the new records get inserted at the first available empty slot (is that actually how this works or is there another row allocation strategy?). Now you go to update those records again...they can't fit in the 1st block...and some percentage of the data won't fit into the 2nd block (depends on the variability in size). So, let's say half the records get relocated....eventually you reach homeostasis. Also...disk fragmentation could be affecting you too but I wouldn't expect an order of magnitude difference on that. Are you on Windows or Unix? Also...are your records indexed for the updates? Michael D. Black Senior Scientist Advanced Analytics Directorate Advanced GEOINT Solutions Operating Unit Northrop Grumman Information Systems _______________________________________________ sqlite-users mailing list sqlite-users@sqlite.org http://sqlite.org:8080/cgi-bin/mailman/listinfo/sqlite-users
