Definitely an excellent response from Alexander, describing the overall state
My suggestion would be to try at least two approaches and develop a way to compare them. Do you want to do statistics? are the entries for those statistics fairly immutable? it seems TS would be a good choice. User verification could also work, with configurable levels of consistency from KV, how much is enough?; no one but the authors of the software can answer for sure. Without fully understanding all the requirement any one could call a solution infantile or overly engineered and not be sure which is correct. I can say that even a bad test will give you something to compare with, but one needs to compare fully.. how will your application respond to sudden server failure?, how will it respond to demand? make a small scenario and basho_bench it, apply more criteria, see where you get backed into a corner.. The best ways to approach solving a problem seem bountiful until you apply all the criteria required of the solution. Less criteria is usually better, Riak has a pretty stringent criteria for a database software, to scale and to remain highly available while making relatively efficient use of resources.. Your designs should always try to complement this. It is often hard to determine you've done that (distributed computing probs) until you have a proof of concept and start beating on it. It seems in one of your use cases "game scores" this is relatively immutable... Users made a score at a given time. Since they would not go back and change this later.. TS could be potentially more efficient if you designed around this. If you tried to design it in literally any configuration you would need to test it and optimally make the test available if you wanted to verify it will work for you over time, or to confer with someone else. Best, Pedram On Thu, Jun 16, 2016 at 12:28 PM, Gianluca Padovani <gpadov...@gmail.com> wrote: > Hi Alexander thanks for your response. > > 2016-06-16 6:19 GMT+02:00 Alexander Sicular <sicul...@gmail.com>: >> >> My question to you is what is your >> use case? >> > > This is the problem, I don't know it clearly. There are some use cases very > clear. For example I need to verify username and password on a big bunch of > users, manage some data about users, and I think that RiakKV is perfect for > this scenario, I clearly identify the user by username or email and check > the password, update some data about it some earned points and some money. I > think that riakKV is a perfect choice. > > Other scenario is to retrieve some data about the user something similar > about your score or your games. This scenario is more difficult because I > need to maintain a relation between the users and this objects, I'm > evaluating to embed it in the user object but I don't think it's a good > idea, because we have a lot of this object and probably I need to create a > special keys or an intermediary objects to get all this data, divided by > days ??!? In this scenario riakKV probably with Solar or 2i is yet good. > > Another scenario is something more statistical/analytical, and I don't know > exactly but something like this: find the first 5 users with the best score > divided for every region and for every city ... In this case, probably, > RiakKV is not a good fit, (map/reduce?!?) > > Any suggestions? :-) > > thanks and bye > Gianluca > > _______________________________________________ > riak-users mailing list > riak-users@lists.basho.com > http://lists.basho.com/mailman/listinfo/riak-users_lists.basho.com > _______________________________________________ riak-users mailing list riak-users@lists.basho.com http://lists.basho.com/mailman/listinfo/riak-users_lists.basho.com
