Admission Control for Distributed Database Transactions
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In distributed databases, transactions competing for IOPS capacity can lead to IOPS starvation and live-lock situations, where transactions are aborted due to lack of resources, causing cascading failures and reduced performance.
Innovation Solution
Implementing a capacity-based admission control mechanism using a bounded queue or data structure that sorts requests by timestamp or priority, allowing only admitted requests to proceed, thereby reducing contention and ensuring sufficient resources for transaction completion.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If transactions are allowed to execute without admission control, then transaction throughput is maximized, but IOPS starvation and live-lock situations occur causing transactions to be aborted
Solution Approach 1:
The patent applies preliminary action by implementing admission control before transactions consume IOPS resources. The system pre-evaluates whether a transaction should be admitted based on future IOPS projections, preventing resource exhaustion before it occurs. This allows the system to maintain high throughput while ensuring reliable transaction completion by avoiding IOPS starvation and live-lock situations through proactive resource management.
2Reliability
If admission control is implemented to prevent IOPS starvation, then transaction reliability is improved, but system complexity increases due to additional control mechanisms
Solution Approach 1:
The patent applies parameter changes by dynamically adjusting admission decisions based on projected IOPS consumption and available capacity. The system monitors IOPS usage patterns and changes admission parameters in real-time, allowing reliable transaction completion while managing complexity through data-driven decision-making rather than complex rule-based control.
Solution Approach 2:
The system implements self-service by having storage nodes autonomously evaluate admission requests based on their own IOPS capacity and projected needs. Each node independently determines whether to admit a transaction without requiring centralized coordination, reducing overall system complexity while maintaining high reliability through localized decision-making.
3Loss of energy
If transactions wait for IOPS capacity availability, then resource utilization is optimized, but transaction latency increases due to waiting periods
Solution Approach 1:
The patent applies preliminary action by pre-admitting transactions before IOPS capacity becomes constrained. The system projects future IOPS needs and admits transactions in advance when capacity is available, ensuring they can complete without waiting. This maintains optimized resource utilization while minimizing transaction latency by eliminating the wait period entirely.
Solution Approach 2:
The system implements dynamics by continuously adjusting admission thresholds and IOPS projections based on real-time workload patterns. The admission control mechanism adapts its behavior dynamically to match actual IOPS availability and transaction characteristics, optimizing both resource utilization and transaction speed by responding to changing system conditions rather than following fixed rules.
Data Source
AI summary
Various embodiments are described regarding admitting for performance ordered operations of atomic transactions across a distributed database. In some embodiments, a node of a distributed system receives and accumulates requests for operations over a period of time. The node can then order the requests based on certain criteria, such as by timestamps of associated transactions. The node can then perform an admission control on the ordered requests, until a capacity of the node to perform the requests for operations is exceeded. During the admission control process, those requests for operations received within the period of time that are not admitted can be rejected by the node.