Just started the development of the successor of my application OnePlaylist. There will be more than one app to be precise but this is the first one and colser to OnePlaylist after pivot. OnePlaylist had more than 4000 visitors a day, so i hope more for the successor on a longuer term: 3 or 4 years versus one year.
Its time to rebuild OnePlaylist without the architectural mistakes related to the business opportunity and cost. This time i will count every bytes put on NoSQL solutions. I will also store data from services like dbpedia.org first on files and then push only data i need on the NoSQL db. I f i need new fields, i will grab them again from disk and push it to the NoSQL store. Indeed i’m going for a super low cost solution this time. I want to minimise RAM usage which is the first waste in a NoSQL world of big data and optimistic business plans. I hope everything will be kept on a single vm this time….
It will on this project my primary store, and unique. I will use redis too but, again, minimise usage and add lot of timeout for keys. I use retire ruby gem on this ruby on rails application. I decided to not use the model class wrapper: in raw mode in best effort, using aliases, simple storing with async task and rake… The cluster setting on sharding and replication will be tune again for a minimum cost cluster.