Fan-out Message Queuing for Distributed Workload Distribution
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Modern computing systems face challenges with workload distribution across multiple hosts, leading to potential correlated failures and uneven task distribution, resulting in increased latency and underutilization of resources due to shared resource unavailability and disproportionate load.
Innovation Solution
A publication-subscription system employing messaging queues to apportion workload across multiple instances, using a fan-out model to sub-divide topics into fragments, allowing each instance to manage a manageable workload and ensure timely message delivery without overloading any single host.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If multiple computing resources are shared to provide services, then resource utilization is improved, but correlated failures occur when a host becomes unavailable
Solution Approach 1:
The patent segments the subscriber base into multiple fragments, where each fragment is assigned to a different instance. This segmentation ensures that if one host fails, only the subscribers in that specific fragment are affected, while other fragments continue to receive messages through their respective instances, thereby preventing correlated failures across the entire system.
2Device complexity
If workload is concentrated on fewer hosts, then device complexity is reduced, but task distribution becomes uneven causing increased latency
Solution Approach 1:
The system segments the workload by dividing subscribers into fragments and assigning each fragment to a separate instance. This distribution of workload across multiple instances prevents any single host from becoming overloaded, thereby maintaining efficient message delivery latency while managing system complexity through the fragment-based approach.
Solution Approach 2:
The patent introduces a new dimension of organization by creating fragments of subscribers rather than distributing individual subscribers directly. This fragment-based dimension allows for more granular control and balanced workload distribution across instances, improving latency without excessively increasing complexity.
3Device complexity
If all subscribers are assigned to a single instance, then device complexity is minimized, but workload becomes disproportionate leading to underutilized resources
Solution Approach 1:
The patent divides the subscriber base into multiple fragments and assigns each fragment to different instances. This segmentation enables proportional workload distribution across multiple instances, ensuring that resources are utilized efficiently without creating disproportionate loads on any single instance, thereby improving overall resource utilization.
Data Source
AI summary
Current distributed message queuing services are designed to share resources in order to fulfill greater amounts of information; however, resource sharing often leads to correlated failures, task distribution problems, and misaligned load balancing of those shared resources. By maintaining a queue of publish request messages to be retrieved by message processors when the workload of the message processors is available, the enqueued messages are processed and fanned-out in order to provide information in the messages to the subscribers without latency or loss of data.