Introduction

For data analysis, it is common to query hot and cold data at varying frequencies and response speed. For example, to analyze user behavior, traffic data must be frequently queried and and the requests need to be quickly responded. Historical data that is seldomly accessed needs to be backed up over a long period for audit and backtracking. Query demands on such data decrease sharply as time goes by. If all data is stored locally, a large number of resources will be wasted.

This topic is available for MRS 3.3.1 or later only.

Principles

Apache Doris 2.0 supports the separation storage of cold and hot data. You can use this function to sink data from the local host to the object storage.

Figure 1 Hot and cold data separation

OBS provides secure, reliable, and cost-effective distributed storage service that supports large-scale data. Doris uses OBS to store data separately. Figure 2 shows the principle.

Figure 2 Cold and hot data separation principle