3 ms·
Key-Value Stores: a practical overview
- SlyShy 17y agoI've started using Redis for the majority of my projects, with one caveat. I still need full-text search, so I use Redis in combination with MySQL. Small and frequently accessed user data is all stored in Redis, while the large chunks of text are kept in MySQL. That way I kept the speed of Redis, and still get to use Sphinx.
- iamaleksey 17y agoYou can use sphinx xmlpipe or xmlpipe2 datasource [http://www.sphinxsearch.com/docs/current.html#xmlpipe2 http://www.sphinxsearch.com/docs/current.html#xmlpipe2] and get rid of the middle-man (MySQL).
- SlyShy 17y agoThanks for the link, that's an interesting possibility. I probably won't be getting rid of MySQL however, because it serves another purpose for me. I'm not comfortable having my entire dataset in memory, because I'm currently constrained in my server setup and I'm dealing with relatively large datasets per user.
- roder 17y agoI was disappointed to not see Riak covered.
- z8000 17y agoI really like riak but I would love to see a couple of things (which are coming after talking to the developers): select/put parts of a document instead of fetching to the client, updating, and sending the whole document back; writing map-reduce in something other than erlang (sue me, I am not an erlang expert yet); performing more operations server side on documents in situ. I am playing with these ideas against a json object store, taking inspiration from riak, redis, and couchdb; based on node.js. It's vaporware ATM but if it pans out I'll open it up.