Bear in mind as well how often you will be updating certain information: each update creates a document revision, so a large document where a couple of fields (or even just one) are frequently updated can lead to increased storage requirements and will also impact replication: there will be frequent replication of a large document where only a small part is actually changing. In a scenario like that, it may well be better to have the frequently updating field(s) in separate documents.
Regards, Andrew On 9 April 2012 17:26, Mohammad Prabowo <[email protected]> wrote: > Thanks! I had read somewhere that there is a tradeoff between embedding the > data (example 1) or more normalized document (example 2). It's more of a > choice between data locality, disk space, and querying flexibilities. I > guess since every query must go trough views, the speed benefits of data > locality is therefore reduced > > On Mon, Apr 9, 2012 at 9:01 PM, Keith Gable <[email protected] > >wrote: > > > I'd go the first route, but salaries and titles should be arrays of > hashes: > > > > "titles": [ > > { "name": "xxx", "from": "xxx", "to": "xxx" } > > ] > > > > If you want to decouple the data, like if you wanted a list of all > titles, > > you'd use CouchDB views. > > On Apr 9, 2012 6:07 AM, "Mohammad Prabowo" <[email protected]> wrote: > > > > > Hi, suppose i have relational db with schema like this > > > > > > employees-schema< > > > http://dev.mysql.com/doc/employee/en/images/employees-schema.png> > > > > > > I want to try converting it into document. I have two question: > > > > > > 1. The main strength of Document is that it is 'self contained'. > > Meaning > > > we don't need to do JOIN stuff, and all data that is needed are > > contained > > > within documents. So, should i choose to use nested documents like > > this : > > > > > > { > > > "emp_no": "...", > > > "birth_date": "...", > > > "first_name": "..", > > > "last_name": "..", > > > "gender": "..", > > > "hire_date": "..", > > > "titles": { > > > "title": "...", > > > "from_date": "...", > > > "to_date": "..." > > > }, > > > "salaries": { > > > "salary": "...", > > > "from_date": "...", > > > "to_date": "..." > > > } > > > } > > > > > > > > > or using different documents like this : > > > > > > [ > > > { > > > "doc_name": "employees", > > > "emp_no": "...", > > > "birth_date": "...", > > > "first_name": "..", > > > "last_name": "..", > > > "gender": "..", > > > "hire_date": ".." > > > }, > > > { > > > "doc_name": "titles", > > > "from_date": "...", > > > "to_date": "..." > > > }, > > > { > > > "doc_name": "salaries", > > > "salary": "...", > > > "from_date": "...", > > > "to_date": "..." > > > } > > > ] > > > > > > > > > 2. I want to benchmark MySQL and CouchDB with > > > YCSB<https://github.com/brianfrankcooper/YCSB/wiki>. > > > Is there are db layer that has been built for CouchDB ? > > > > > > Thanks in advance > > > > > >
