问题
I am looking to build a realestate search engine specs are Approx 500 000 listings daily updates of potentially 50 000 listings Data supplied in clean(ish) CSV's - need to remove characters, encode utf, the usual. 50+ fields of data (30 images, various property specs etc)
Im having a lot of problem with Drupal7 and Joomla cannot handle it. Thats just the data import.
Im wanting to have solr index the data and serve as the search engine. I have a few questions.
- Can solr serve the listing directly from its index? (If so do I need a data store such as Mysql or even a CMS)
- Would I be better off putting the data in a simple single table mysql DB and use that to push documents to solr for index, then either load listings from the DB or from Solr index.
Due to data difficulties, it seems I can simply do away with a lot of complications trying to figure out the inner workings of D7/Joomla/any other cms and just put a few simple php files up as the front end.
I dont need anything fancy looking, was going to use the basic drupal template for this project.
I need speed and reliability and excellent search results.
回答1:
IMHO it should be possible to use SOLR exclusively for your purpose. The number of 50000 listings is not very much for SOLR even for a single server, but 500000 updates per about 10h I suggest is indeed a lot. Since you will have about 50000 updates per hour which is equivalent to a full reindex per hour.
We use SOLR for our enterprise, too, and with something about 40-120 fields. 40000 items do need about 5 minutes to index completely. If you want to autowarm caches you have to add perhaps some minutes to that.
As far as I see your problem will be the small update periods. If you want to update individual documents instead of all of the 50000 listings once per hour, your solr cannot use caching or you will have to use multiple solr servers. (Perhaps for solr 4.0 you could even consider scaling up your solr server hardware, but i suspect 3.x would have any benefits from that) No use of caches could lead to slower search performance, but it does not have to.
Since SOLR offers thy dynamic fields functionality you can add different structures per document. This should match your various properties requirement.
来源：https://stackoverflow.com/questions/10052453/can-solr-serve-individual-documents-like-a-cms-would