incremental reindexing on texis/linux

[jkj2001]

We're using your linux version of Texis, version 4.04.1067366033, and wondered about fragmentation and speed issues if we were to incrementally reindex (i.e. not drop the indexes first) our database.

At one time we used your Windows NT product and incremental reindexes slammed our CPUs and gave us index heartburn, in that there was a fair amount of fragmentation, temp files and a bad time all around. Ever since we've gone the drop/recreate route when it comes to reindexing.

However...we'd like to explore the feature again. One of our databases is 35 GB in size and changes fairly little day to day, just enough to bloat the indexes a little more than we'd like. Would it be fair to say incremental reindexing works better in Linux than on Windows NT, and if so, what tips or flags would you reccomend? Of course we would schedule fresh indexes periodically; in no sense would we try "skating by" on incrementals forever. Thanks!

[John]

When the incremental indexing is done the index state is identical to a drop and fresh index, so there shouldn't be a need from a file content point of view to drop and reindex.

It sounds as if the issue was due to the filesystem handling and fragmneting in the operating system. Whether Linux will do better will depend on which filesystem you are using and how much free space is on the filesystem.

Having plenty of free disk space may help, alternatively having two filesystems for the index, and on each incremental index switch filesystem should help eliminate any fragmentation issues.