"Recommended System Introduction" NetFlix-System Overview

Reading background:

1: What do you need to know about Netflix?

2: The basic concept of recommendation system?

3: Real-time recommendation system.

Purpose of reading:

A precise understanding of the hierarchical structure of the current recommendation system: to establish a fine division between the various levels.

system is divided into3section,Online,Nearlineand Offline sections(OffLine). Onlineand theNearlinepart is considered to be the real-time part.

The offline part is the main body of the traditional personalized recommendation system, and periodically takes out a large amount of stored data for batch operation, then establishes and updates the model, which is not described in detail here.

nearline memcached, Cassandra MYSQL etc can be quickly queried in the store as a backup. In the netflix in the system, their flow calculation is through a called netflix.manhattan< Span style= "FONT-SIZE:12PX; Font-family: the song body; > to achieve, it is a similar to the storm Real-time streaming computing framework, just for their own applications have a certain specificity, is not a universal real-time computing framework.

online part. This section takes advantage of the offline part of the body model and takes into account the nearline

The key problem of personalized recommendation system architecture is how to realize the process of online and offline computing in a seamless way, in the final analysis, it is also the design of the algorithm, how to distribute the calculation steps of the algorithm reasonably to the various parts, so that the model can be very accurate and can be calculated quickly to meet the requirements of real-time.

Picture exceeds upload limit: pending:

"Recommended System Introduction" NetFlix-System Overview