Looking for Riak recommendations for modeling data with N:1 references

elij elij.mx at gmail.com
Fri Jun 8 01:16:17 EDT 2012


On Monday, June 4, 2012 at 2:53 PM, Mark Phillips wrote:
> Hi, 
> 
> Assuming I'm understanding your use case correctly, the first option seems like your best bet. 
> 
> Primary key lookups are where Riak really shines, and if you're only GET-ing one key and then manipulating the associated json-dict with some app code, things should perform quite well (assuming your objects aren't huge). 
> 
> Does that make sense? Or did I misunderstand things?
I think so. I am basically trying to find out what is faster for largish buckets (hundreds of millions of json blobs, but each relatively small -- around 500 to 700 bytes or so):

* single level link walking
* secondary indexes where there is only ever one possible result for a given index value
* a separate bucket where a store a key as a value/'pointer', which would correspond to another key in another bucket (basically a self-roll of one of the above).





More information about the riak-users mailing list