Looking at Hadoop, the distributed file system. I don;t see anything in a distributed file system that Json bots can handle. Hadoop is designed for very large datawarehouses, and it is a method to spawn small client processes to gather data from clusters of Datanodes, guided by a NamedNode. They have two protocols, one for the client and one for the server to datanodes.
But, using the Fresno desgn, Json queries are easily duplicated and spun out to other nodes, collectibng results in proper order and delivering a Json. The Fresno system still needs a distributed indexing system. The Fresno design assumes Json have self directing capability, once at the right node they know how to extract data with query by example.
All of these systems, ultimately rely on spawned child processes to intelligently grab data. The difference is the Fresno Design intends everything to be a binary operation of two Jsons resulting in Json out. Can the industry ultimately define a lightweight Json for this? Yes, and they will do it. For myself, I am still uninterested in stepping through all this with a bunch of industry baggage to deal with. In other words, a little software now and then, but mostly just watching the market to see if they can get a Json that works.
I am hoping for a graph theorists to get going on this, actually. But eventually, if one doesn't show up, I will work more details.
No comments:
Post a Comment