Workload-Aware Database Selection for Balanced Resharding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing database resharding techniques that split servers based on size often result in uneven workload distribution, as some databases may have higher input/output operations regardless of size, leading to inefficient resource utilization.
Innovation Solution
Analyze metrics such as reads and writes to compute weighted sums for each database instance, applying different weights to account for varying impacts, and divide instances into groups for an even split in workload.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If databases are split based on size for resharding, then the data distribution appears balanced, but the workload distribution becomes uneven
Solution Approach 1:
The patent changes the selection criteria from a single parameter (database size) to multiple parameters including read operations, write operations, and connections. By computing a weighted sum of these parameters, the system achieves balanced workload distribution rather than just balanced data distribution, resolving the contradiction between data balance and workload efficiency
Solution Approach 2:
The patent creates a composite metric by combining multiple workload parameters (reads, writes, connections) with different weights into a single selection criterion. This composite approach allows the system to evaluate databases holistically and select those that will achieve overall workload balance across all resource types, not just data size balance
2Quantity of substance
If vertical scaling is used to increase server capacity, then the database capacity increases, but the backup or cloning time increases significantly
Solution Approach 1:
The patent divides the single large database server into multiple smaller database instances across multiple servers through resharding. This segmentation allows individual instances to be backed up or cloned more quickly while maintaining overall system capacity, thus resolving the contradiction between capacity and backup time
Solution Approach 2:
The patent performs resharding proactively before vertical scaling becomes necessary, distributing databases across multiple servers in advance. This partial distribution allows the system to maintain capacity while enabling faster backup operations on individual smaller instances rather than one large instance
Data Source
AI summary
Methods, systems, and devices for data management are described. A database server may identify workload metrics for a set of database instances and generate a weighted sum of the workload metrics for each database distance. The database server may select, for a resharding operation, a first set of database instances and a second set of database instances, and the selection may be based on a relationship between a first sum of one or more respective weighted sums for the database instances in the first set and a second sum of one or more respective weighted sums for the database instances in the second set. The database server may execute the resharding operation that results in the first set continuing to be supported by the first database server and the second set being supported by the second database server.