Thanks for your input (also the_duke)! If time permits, we may come up with a second blog post on this topic.

If I understood your example query, your application serves vendors and each vendor has different products. Is that correct?

You can approach this sharding question in one of two ways.

1. Merge different product tables into one large product table and add a vendor column

2. Model product tables as "reference tables". This will replicate the product tables to all nodes in the cluster

Without knowing more about your application / table schemas, I'd recommend the first approach. I'd also be happy to chat more if you drop us a line.