Impact of Latency on Database Replication
Databases store valuable data, such as the amount of money in someone’s bank account, orders placed on a webshop, or other valuable customer data. It is important that databases which store this data perform optimally, have low response times, make the latest data available to their users, and remain consistent. An often-used best practice to achieve these characteristics is to replicate data across multiple nodes and serve this data from geographically separated locations.
Database replication is a mechanism used in distributed systems, as replication adds fault-tolerance and high availability characteristics to a database system. As more businesses expand across the globe, multi-availability zones are used to deploy their infrastructure in different regions using cloud platforms. Database replication aids in keeping data consistent across these different multi-availability zone regions.
Network latency in the replication process
However, because the data is replicated over different regions in the world, the distance between these zones introduces network latency in the replication process. Latency is the most significant component of the total round-trip time (RTT) in high-speed networks, outlined by. Depending on the consistency model and the replication setup used, latency can impact aspects such as database performance, data availability, and data consistency. Latency therefore introduces complexity in the form of trade-offs depending on the replication and consistency model used in an implementation.
This research aimed to uncover the impact of network latency on the process of transactional database replication in simulated geographically separated database environments.