Virtual Stand-In Service for Production Computing Outages
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Production computing services face significant expenses due to the need for redundancy and hot backup systems, which can be beyond the control of clients, and experience outages that lead to downtime, disrupting service availability.
Innovation Solution
A virtual stand-in computing service is generated based on transaction data, including multiple request types and confidence data, which allows it to respond to client requests during outages by using a transaction library to provide model-generated responses and update the production computing service model upon availability, ensuring data parity and adaptive learning.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If complete redundancy with parallel production computing services is provided, then service availability and reliability are improved, but operational costs double
Solution Approach 1:
The patent creates a virtual stand-in service that copies only the essential behavior and response patterns of the production service through transaction sampling and model generation, rather than maintaining a complete redundant system. This allows the stand-in to handle requests during outages without requiring full operational capability, significantly reducing costs while maintaining reliability.
Solution Approach 2:
The system performs partial action by capturing and processing only a sample of transactions to build the service model, rather than replicating all service operations. The virtual stand-in service handles only enough functionality to maintain service continuity during outages, using confidence scores to determine when to escalate to full production service restoration.
2Reliability
If hot backup systems are implemented for production computing service, then service continuity is improved, but system complexity and control difficulty increase
Solution Approach 1:
The virtual stand-in service acts as an intermediary between clients and the production service during outages. It receives client requests, uses the generated model to determine appropriate responses, and only contacts the production service when needed for model updates or when confidence thresholds are exceeded, simplifying the backup architecture.
Solution Approach 2:
The stand-in service is self-sufficient during outages, using its generated model and confidence scoring to autonomously handle requests without requiring active management or coordination with the production service. The system automatically updates its model when the production service becomes available again, reducing operational complexity.
3Loss of energy
If virtual stand-in service uses transaction sampling to generate service model, then cost and complexity are reduced, but service accuracy and reliability may be compromised
Solution Approach 1:
The system continuously monitors the accuracy of model-generated responses by comparing them with actual production service responses when available. This feedback is used to update the service model and adjust confidence scores, ensuring that the sampled transaction data accurately represents the production service behavior while maintaining cost efficiency.
Solution Approach 2:
The system dynamically adjusts confidence thresholds and sampling parameters based on service characteristics and outage conditions. By changing these parameters, the system can maintain acceptable service accuracy while optimizing the balance between model generation costs and response reliability.
Data Source
AI summary
Provided are methods of providing a virtual service that may provide partial real time service to clients of a production computing service that is unavailable. Methods may include generating, based on transaction data corresponding to a production computing service that is available, a production computing service model that includes multiple request types and multiple confidence values that correspond to ones of the request types. Methods may include responding to a request received from a client of the production computing service with a model-generated response to the request in response to the production computing service being unavailable. The production computing service is updated with the request received from the client and the model-generated response responsive to the production computing service being available after being unavailable.