Cloud Transaction Scheduling via Preliminary Resource Coordination
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In cloud-based systems, the execution of global transactions is inefficient due to the need for multiple microservices to complete local transactions, where failure of one microservice results in resource wastage and rollback of other transactions, leading to reduced system efficiency and performance degradation.
Innovation Solution
A transaction management system with schedule managers for global and local transactions that ensures resources are available for transaction execution by coordinating the reservation and confirmation of resources across participant services, thereby preventing unnecessary resource allocation and rollback.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If multiple microservices execute local transactions in parallel for a global transaction, then transaction completion speed is improved, but system reliability deteriorates due to potential failures and rollbacks
Solution Approach 1:
The patent applies preliminary action by having the transaction coordinator send out transaction requests to all participant microservices before any actual resource allocation occurs. The coordinator waits for all participants to confirm their readiness and resource availability before signaling them to proceed with the transaction. This ensures that if any participant fails, no resources have been allocated yet, eliminating the need for rollbacks while maintaining parallel execution capability.
2Reliability
If microservices reserve resources early for local transactions, then resource availability is improved, but resource wastage increases when transactions fail and require rollback
Solution Approach 1:
The system performs preliminary resource reservation confirmation without actual resource allocation. The coordinator requests resource availability information from all participant microservices in advance, and only after receiving confirmation from all participants does the system proceed with actual resource allocation. This eliminates resource wastage because resources are only reserved when the entire global transaction is guaranteed to succeed, while still ensuring resource availability through advance planning.
3Manufacturing precision
If the system waits for all microservices to complete their local transactions before proceeding, then transaction accuracy is improved, but system productivity deteriorates due to sequential processing
Solution Approach 1:
The transaction coordinator sends requests to all participant microservices simultaneously in parallel, rather than waiting for each to complete sequentially. Each microservice executes its local transaction independently and concurrently. The coordinator collects results from all participants and determines overall transaction success based on all results. This maintains transaction accuracy by verifying all participants while dramatically improving productivity through parallel execution.
4Productivity
If the system allows microservices to proceed with local transactions without prior coordination, then system efficiency is improved, but resource allocation accuracy deteriorates leading to conflicts and rollbacks
Solution Approach 1:
The transaction coordinator serves as an intermediary between all participant microservices. It sends transaction requests to all participants simultaneously, collects their resource availability information, determines whether all resources can be allocated, and then signals all participants to proceed or abort. This intermediary coordination ensures resource allocation accuracy by verifying all participants' readiness before allocation, while maintaining system efficiency through parallel processing and avoiding sequential bottlenecks.
Data Source
AI summary
Methods, systems, and computer-readable storage media for receiving, by a scheduled transaction manager, a first request for a first global transaction for an application executed within a cloud-based system, the application including a set of services where execution of the first global transaction requires a set of participant services, in response to receiving the first request, transmitting, by a scheduled transaction coordinator, a first set of requests for a first set of local transactions to the set of participant services, receiving, by the scheduled transaction coordinator, indications of reserved resources from participant services, and determining that received indication of reserved resources have been received from all participant services in the set of participant services, and in response, inhibiting cancelation of resource reservations for each of the participant services in the set of participant services, and receiving one or more results of local transactions in the set of local transactions.