Sharding Use Cases

What specific use cases do we want to address with db partioning (and other techniques) that are challenging to scale? List here for discussion.

video site (e.g., youtube) (also, GridFS scale-up)
- seems straightforward: partition by video
- for related videos feature, see search below
social networking (e.g., facebook)
- this can be quite hard to partition, because it is difficult to cluster people.
very high RPS sites with small datasets
- N replicas, instead of partioning, might help here
  - replicas only work if the dataset is really small as we are using/wasting the same RAM on each replica. thus, partioning might help us with ram cache efficiency even if entire data set fits on one or two drives.
twitter
search & tagging

Log Processing

Use cases related to map-reduce like things.