Storage Server Selection by Request Bands for Independent Throttling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Cloud computing file store servers face performance issues due to throttling when a certain percentage of threads is busy, affecting service level agreements and overall system performance, especially with increasing incoming requests.
Innovation Solution
Implementing a system that classifies requests into bands based on characteristics such as size, response time, and organizational characteristics, allocating computing resources differently to each band, and using traffic lights to manage request processing within each band independently.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If the file store server throttles requests when a certain percentage of threads is busy, then the server protects itself from overload, but the service level agreements are not met and system performance deteriorates
Solution Approach 1:
The patent segments requests into different bands (e.g., read requests, write requests, bulk operations) based on their characteristics. Each band is then managed independently with its own throttling mechanism, allowing the server to process different types of requests simultaneously without being blocked by a single band's throughput limitations.
Solution Approach 2:
The patent applies different quality levels of service to different request bands. Critical bands (such as read requests) may receive higher priority and less aggressive throttling compared to non-critical bands (such as bulk write operations). This local differentiation ensures that service level agreements are met for essential operations while still managing overall server load.
2Reliability
If the application server waits until the traffic light is reset to send requests, then the file store server is protected from overload, but the overall usage of customer applications is negatively affected
Solution Approach 1:
By segmenting requests into multiple bands with independent traffic lights, the system allows different bands to be processed at different times without blocking the entire request queue. This reduces the perceived wait time for individual requests while maintaining server stability.
Solution Approach 2:
The system performs preliminary classification of requests into bands before they reach the file store server. This preliminary action allows the application server to prepare and queue requests in an organized manner, reducing the time spent waiting for the traffic light to reset and improving overall response time.
3Device complexity
If all requests are treated equally and processed in a single queue, then the system is simple to manage, but resource utilization is inefficient and larger requests block smaller requests
Solution Approach 1:
The patent divides the single request queue into multiple band-specific queues (e.g., read queue, write queue, bulk operation queue). Each queue is managed independently with its own traffic light mechanism, allowing parallel processing of different request types and improving resource utilization without requiring complex centralized management.
Solution Approach 2:
The system applies partial throttling to specific bands rather than throttling all requests uniformly. This allows the server to process critical requests (reads) while limiting non-critical requests (bulk writes), improving overall resource utilization efficiency without implementing overly complex discrimination mechanisms.
Data Source
AI summary
In some embodiments, a method receives a request for accessing a database and selects a band from a plurality of bands for the request based on a characteristic of the request. A band is associated with a set of characteristics. The method selects a storage server for the band and determines whether the storage server can process requests from the band. When the storage server can process requests from the band, the method causes processing of the request by the storage server to access the database. When the storage server cannot process requests from the band, a traffic limiting indication is set for the storage server to limit processing of requests for the band on the storage server.