Hbase and traditional database

NoSQL is too hot, and there are too many products, which are conservatively estimated to be hundreds.

Internet companies are generally concentrated in the following categories, and each category only gives one more common or successful example.

1. KV storage in memory: Redis

In memory, key-value storage provides richer data structure and computing power, successfully replaces memcached, provides fast downtime recovery through checkpoints and log submission, and supports replication to provide read scalability and high availability.

2. Disk-based KV storage: Leveldb

The real key-value storage based on disk has the advantages of simple model, unlimited data size and high data reliability. The excellent products produced by several great gods of Google, the LSM mode is naturally optimized, and the sequential disk writing method is perfect for the new hardware ssd. The disadvantage is that only one library is provided, and the server needs to package it itself.

3. Document storage: Mongodb

Distributed nosql has the biggest highlight different from mysql: scalability. The latest attraction of mongodb is that it provides sql interface, which is the most mysql-like interface in nosql at present, but it has no characteristics of ACID, fast development speed, index support and other characteristics, and is easy to use. For scenarios where the amount of data far exceeds the memory limit, you need to be cautious.

4. list storage: HBase

This rich second generation doesn't seem to need to go into details. The biggest advantage is open source. For basic queries such as ordinary scanning and row-based get, performance is not a problem at all. Only bare api is provided, which is short-board in ease of use and has the strongest expansibility. Secondly, the community has caught the express train of Hadoop and developed rapidly. There are many open source products based on Hadoop to solve complex queries such as join and aggregation operations.