Decoupled Commit Rate Management for Cloud Data Consistency
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing cloud computing systems face challenges in maintaining consistency across data operations between front-end and middle tier servers, particularly due to the complexity of implementing logic for commit operations on backend storage servers, which leads to difficulties in programming and separating services across these tiers.
Innovation Solution
Decoupling commit operations from write operations and using tracking objects to manage commit rates, allowing for dynamic adjustment based on measured latency to optimize latency and ensure consistency, while utilizing a generic scale-out store with a dynamic control loop for low latency and fairness.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If commit operations are coupled with write operations and implemented through complex logic on backend storage servers, then data consistency can be maintained, but the system complexity increases and programming becomes difficult
Solution Approach 1:
The patent segments the commit operation from the write operation, creating independent tracking objects for commit operations. This separation allows write operations to be performed independently while commit operations are tracked and executed separately, reducing the complexity of implementing consistency logic across the entire system.
Solution Approach 2:
The patent introduces tracking objects as intermediaries between write operations and commit operations. These tracking objects serve as mediators that record write operations and manage their commit status, simplifying the implementation of consistency semantics by providing a standardized interface for tracking and committing changes.
2Reliability
If commit operations are performed frequently to maintain consistency, then data consistency is improved, but latency increases
Solution Approach 1:
The patent implements periodic commit operations where tracking objects are provided to computing devices at a defined commit rate rather than immediately after each write operation. This periodic approach allows the system to balance consistency requirements with latency optimization by adjusting the commit rate according to measured performance.
Solution Approach 2:
The patent dynamically adjusts the commit rate based on measured latency. The system monitors the latency associated with commit operations and compares it to the defined commit rate, then adjusts the commit rate accordingly. This dynamic adaptation allows the system to optimize latency while maintaining appropriate consistency levels.
3Reliability
If commit rate is increased to improve consistency, then data consistency is improved, but system performance and responsiveness deteriorate
Solution Approach 1:
The patent implements a feedback mechanism where the system measures latency associated with commit operations and uses this measurement to adjust the commit rate. The measured latency is compared to the defined commit rate, and the commit rate is adjusted based on this comparison and a factor determined relative to the defined commit rate. This feedback loop enables the system to automatically optimize performance while maintaining consistency.
Data Source
AI summary
Consistency is managed among data operations by coalescing commit operations, adjusting a commit rate to optimize latency, and converging to fairness across servers. Write operations representing a change to data stored by one or more of a plurality of computing devices are received in succession by, for example, a cloud computing service. A state of a plurality of tracking objects is adjusted to indicate whether the change in data has been provided to the computing devices and whether the change in data has been acknowledged by the computing devices. Changes to the same data are coalesced such that only the most recent change is provided to the computing devices for storage. In some embodiments, the commit rate is adjustable such that the commit rate decreases quickly but increases slowly based on measured latencies.


