Thank you all! Sorry, I think 'consistent hashing' is wrong word.
For my use case, I need to store this 'prefix' (either hashed/not) into another table. Will this murmur hashing guarantee next time same string will map to same bytes? And no collision for around 2^10 records? Mingtao Sent from iPhone > On Jul 21, 2014, at 10:28 PM, Ishan Chhabra <[email protected]> wrote: > > Mingtao, > If I understand correctly, you want to prefix the key with a hash (as > mentioned in the book) to get a good distribution. Use MurmurHash (there is > an implementation in HBase code itself) as it is fast and gives a uniform > distribution. > > "Consistent Hashing" is not the correct term to use here if I understand > your intent correctly. > > >> On Mon, Jul 21, 2014 at 2:44 PM, Liam Slusser <[email protected]> wrote: >> >> MD5 isn't a consistent hashing algorithm. Consistent hashing is a scheme >> that provides a hash table functionality in a way that the adding or >> removing of one slot does not significantly change the mapping of keys to >> slots. With that said, a lot of consistent hashing algorithms USE >> md5...but it alone won't get you all the way there. >> >> Some light bedtime reading: >> http://en.wikipedia.org/wiki/Consistent_hashing >> >> liam >> >> >> On Mon, Jul 21, 2014 at 7:18 AM, Mingtao Zhang <[email protected]> >> wrote: >> >>> Hi, >>> >>> I am trying to find a consistant hasing algorithm for the first portion >> of >>> the row key. >>> >>> I saw the document/book that MD5 is mentioned everything. >>> >>> But I have trouble to persuade myself that MD5 ( >>> http://en.wikipedia.org/wiki/MD5) is considered as consistant hasing. >>> >>> Could any of you point me to the library contains the hashing you are >>> using? >>> >>> Thanks in advance! >>> >>> Best Regards, >>> Mingtao > > > > -- > *Ishan Chhabra *| Rocket Scientist | RocketFuel Inc.
