Are there any practical datasets that, by their nature, cannot be partitioned in any practical way?

I am wondering about this because such datasets would be ill-suited to scaling out on commodity hardware and databases containing such data could be forced to scale up on expensive servers, at great cost.

Edit: by nature, I mean that the lack of partitionability is a consequence of the data and of the queries that must be made against it, rather than poor design choices.

  • Any data set which only contains values which are very similar, or has many identical values wouldn't partition very well (e.g., Y/N, 1/0, True/False, NULL/NOTNULL flags). Commented Jun 27, 2014 at 8:55


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Browse other questions tagged or ask your own question.