Function Proxy Switching for Accelerator Failure Continuity
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing techniques for handling accelerator failures in computer systems require restarting or migrating the entire application/VM, leading to gaps in continuous arithmetic processing during failure recovery, sudden failures, and planned switching, which are not addressed by current methods.
Innovation Solution
A signal processing resource switching device with a function proxy execution unit, accelerator failure detection, and offload destination determination unit to automatically switch calculation resources without application modification, ensuring continuous arithmetic processing by detecting failures and predicting potential failures to switch resources seamlessly.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If the entire application/VM is restarted or migrated when an accelerator fails, then the application can continue processing, but arithmetic processing is interrupted during the restart/migration period
Solution Approach 1:
The system segments the application from the accelerator resource, allowing the application to remain active while only the failed accelerator component is replaced. The function proxy execution unit maintains the application's execution context, while the offload destination calculation resource determination unit switches to a backup accelerator, thus avoiding complete application restart and minimizing processing interruption.
Solution Approach 2:
The function proxy execution unit acts as an intermediary between the application and the accelerator. When an accelerator fails, the intermediary redirects function calls to a backup accelerator without requiring the application to be restarted or migrated, thereby maintaining continuous arithmetic processing while ensuring application availability.
2Reliability
If manual detection and switching of accelerators is implemented, then resource switching can be controlled, but the switching response time is delayed and arithmetic processing is interrupted
Solution Approach 1:
The system implements automated feedback mechanisms where the function proxy execution unit continuously monitors accelerator status and automatically triggers switching to backup resources upon detecting failures. This eliminates manual detection delays and ensures rapid response to maintain processing continuity.
Solution Approach 2:
The system performs preliminary actions by pre-configuring backup accelerators and establishing function proxy execution units before failures occur. When an accelerator fails, the switching mechanism is already in place and can immediately redirect processing, minimizing response time and maintaining processing continuity.
3Ease of repair
If accelerators are switched during planned maintenance, then the maintenance can be performed, but arithmetic processing is interrupted during the switching period
Solution Approach 1:
The system applies local quality by maintaining different operational states for different components. While one accelerator is being maintained, the function proxy execution unit continues to route arithmetic processing to other functional accelerators. This allows maintenance to proceed on one component without interrupting overall arithmetic processing productivity.
Solution Approach 2:
The system ensures continuity of useful action by maintaining active arithmetic processing on available accelerators during maintenance periods. The function proxy execution unit continuously redirects processing to operational accelerators, ensuring that productivity is maintained while maintenance activities are performed on individual components.
Data Source
AI summary
A signal processing resource switching device includes a function proxy execution unit configured to accept a “function name⋅argument” from an application unit and notify the application of argument data of a function when the function is executed or ended by a calculation resource, an accelerator failure detection unit configured to detect a failure of an accelerator, and an offload destination calculation resource determination unit configured to determine an unfailed and available resource among the calculation resources, and the function proxy execution unit performs offloading on the resource determined by the offload destination calculation resource determination unit.


