In a distributed system where workload is distributed to multiple nodes, two of the ways of dealing with race conditions where multiple requests to operate on the same data concurrently are the use of consistent hashing and distributed locks. Consistent hashing would ensure that all requests to operate on one set of data are sent to the same worker and distributed locks would ensure that only one worker could operate on any set of data at a time.
My question is what are the pros and cons of either approach and which might be favorable?
The consistent hashing is much easier to implement than a distributed lock. The problem is specific distribution of inputs could be sent only to a subset of the nodes, resulting in some words working harder than others. Distributed lock is harder to implement and requires several communication messages (or some shared data) but won't result in a bias of node allocations.