Cloud Application Load Determination via Queue Wait Indicator
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current cloud computing systems face challenges in determining application load, particularly for advanced applications like packet processing and telecom services, as existing metrics such as CPU utilization and queue length are not suitable, and scaling based on these metrics can be resource-intensive and inefficient.
Innovation Solution
A method is introduced to determine application load in cloud computing environments by using a wait indicator that measures the relation between empty and non-empty states of input queues, allowing for scalable application instances through a load determination component, which can scale applications based on the calculated load.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If CPU utilization is used as a metric for determining application load, then scaling can be implemented for simple web applications, but it is not suitable for packet processing applications that use polling techniques resulting in constant 100% CPU load even when idle
Solution Approach 1:
The patent introduces a wait indicator as an intermediary metric that mediates between the application's actual processing load and the scaling decision mechanism. This wait indicator, derived from queue empty state observations, serves as a more accurate proxy for true application load than CPU utilization, particularly for polling-based packet processing applications where CPU remains busy even during idle periods.
2Measurement precision
If queue length is used as a metric for determining application load, then direct measurement of processing backlog is achieved, but querying queue length is resource-intensive and requires multiple system calls
Solution Approach 1:
The patent extracts the essential information needed for load determination (whether the queue is empty or not) without taking the resource-intensive approach of querying the actual queue length. By observing only the empty state of the queue and deriving a wait indicator from this binary observation, the system achieves sufficient load measurement precision while minimizing computational overhead and avoiding multiple system calls.
3Productivity
If the number of application instances is increased to handle peak load, then system capacity is improved, but resource allocation efficiency decreases when load drops back to normal levels
Solution Approach 1:
The patent implements dynamic scaling of application instances based on the wait indicator metric. The scaling mechanism dynamically adjusts the number of running instances according to actual application load conditions, increasing instances during peak periods when the wait indicator indicates high load and decreasing instances during normal periods when the wait indicator shows low load, thereby optimizing both system capacity and resource efficiency.
Data Source
AI summary
A technique for determining of a load of an application in a cloud computing environment is disclosed. The application is executed with one or more application instances in the cloud computing environment, wherein each of the one or more application instances obtains input data from a respective input queue. A method implementation for supporting the technique comprises determining a wait indicator for at least one of the one or more application instances, the wait indicator for an application instance indicating a relation between empty states of the input queue of the application instance and non-empty states of the input queue of the application instance, and triggering forwarding of the wait indicator determined for the at least one of the one or more application instances to a load determination component.