Hello, I am currently considering Hibernate Shards for horizontal partitioning since we are already using Hibernate throughout our system. I have a few questions / need for some input regarding choice of strategies.

I am thinking about a rather straightforward way of sharding the data based on ID, lets say shard00 consists of entries with IDs = 1 - 15 000 000, shard 01 with IDs 15 000 001 - 30 000 000 etc, using mysql and configureing each db / schema this way and using native ID generation.

My question is then how to solve the shard selection strategy in a good way that allows for easy scaling... I guess one way would be to say that each shard has 15 000 000 entries and then keeping some current max ID to determine on what shard the next entry should be stored by taking currentMaxID / entriesPerShard... But how should this be done in a robust manner? Should we use a support table for keeping track of the highest ID? (sort of like table HI-LO does)? Or is this not a good way of solving this? This way I guess the only thing that has to be changed when expanding with more shards is the configuration file for the new shard and the loading of the session factory. Or?

And I also assume that resolution strategy would then be straightforward aswell since you can determine the shard based on the ID...

Sorry for the fuzzy questions. Hope to get some initial pointers on this from one of you guys that have some hands on experience with shards:)

Cheers,
Aleksander M. Stensby

been thinking more on this and since there are no replies I'll share my (possibly) silly thoughts..

Generating native id's should be no problem, but as far as I can see, my problem lies in the ShardSelectionStrategy... How can I make my strategy aware of the shards when the ID of a new object is not yet known at the stage of shard-selection? My thoughts in the previous post was to use a support table (such as the hibernate_unique_key table) residing on the control shard, and then simply keep track of the current highest ID.. but then I would need access to the control session in the ShardSelectionStrategy, right? And I don't really see how I can supply the control session to the strategy since the strategy is defined and supplied to the strategy factory before the actual shard session factory is created...

I might missing something really important and possibly simple here, so please correct me if I'm off here...

Hoping for some good feedback on this:)

Cheers,
Aleksander