We have a project where we are using TREX to do high volume segmentation with roughly 6 million customers.
We are using infoset queries with an average size of the customer number. But we have 2 problematic tables which are huge. the first one is 45 million records and the second one (and the really important one) is about 640 million records. The TREX server sizing is quite big as well so I'm not inclined to look into hardware before making sure that everything is fine with the software side.
Our problem is the indexing speed of these tables. At the moment our TREX system indexes about 35 million records in 24 hours. That means we need around 3 weeks to index the whole big table with 640 million records.
Is there a way to make the indexing quicker? We don't have much TREX experience so this question might sound quite funny...
I've checked some notes about turning off text mining on the TREX configuration files but it didn't work. Any new created index gets the Text Mining check in the TREX Admin screen.
Help documentation mentions the use of queue server but then it says in CRM segmentation scenario it should be left as default where it is not active.
Any help would be extremely appreciated as this is becoming a showstopper for an important project.
Thanks in advance