Recently I have been reading the debates on SQL vs NoSQL and wanted to get your opinions on the fit/no-fit scenarios with Arc. Currently I store some data records in files, such as the arc forum does, in other projects I am storing data in MySQL. Here are my thoughts/questions: 1. YC uses files as a data storage system for HN correct? I don't know the stats, but I'm guessing they have 10k to 50k user accounts and get at least 10k hits per hour. How does using the file storage system HN/arc uses stack up against SQL & NoSQL? Do you think HN will need to switch with more growth? 2. I plan to spend a few hours to see if I can get couchDB up and running against arc over the next weekend (if I can finish off setting up nginx). I think couchDB should be easy given the data is passed via js over HTTP. Anyone have feedback on couchDB? 3. A common statement is that NoSQL is not ACID compliant thus data integrity is potentially a problem. I have a hard time adjusting to the idea that data integrity is questionable for anything I code. I don't understand the boundaries for when this potential data loss would happen vs not happen. If I house my data on a single server does this mean it will never happen? Is this only a case when syncing these 'data islands'? What are the odds of data loss? Is this a rare thing or a common thing? |