NoSQL Database: Data Consistency

NoSQL Database: consistent reading of data

• High Consistency
  Access any node in the cluster at any time, and the data obtained is consistent;

• User consistency
  For the same user, the data obtained during cluster access is consistent;
  Solve user consistency: Use a sticky session to bind the session to a specific node for processing;
  This will reduce the performance of the Load balancer;

• Final consistency
  Temporary data inconsistency occurs between nodes in the cluster due to delayed data synchronization. However, after data synchronization is completed, the data consistency is achieved;

Update consistency

• Pessimistic Mode

  The use of write locks significantly reduces system response capabilities and may lead to deadlocks.

• Optimistic

  Conflict occurs first, and then the processing method of automatic merging of sequence is extremely "domain-specific ".

Relax "consistency constraints"

• CAP Theorem
  Consistency, Availability, and Partition tolerance. The three attributes can only satisfy two at the same time;

Partition tolerance explanation: the cluster is still available when it is divided into multiple due to communication faults

• CA system

  "Partition" is not available in a single server cluster

• PA/PC

  When a "partition" occurs in a cluster, the trade-off between "consistency" and "availability" generally sacrifices partial consistency (eg: Use final consistency) to ensure availability.

Relax "persistence" constraints

More Strict Persistence means more performance loss;

• Sacrifice "durability" in exchange for better performance
• Replication "persistent" Fault

  Master node failure, not synchronized to data loss from the slave node master node recovery, data conflict solution updated during the fault: specify the required durability for a single request

Mind Map

Reference

NoSQL Essence

Posted by: Large CC | 02JUL, 2014
Blog: blog.me115.com [subscription]
Weibo: Sina Weibo