Task Type Determination for Auto-Scaling in Distributed Systems

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing techniques fail to accurately determine whether Structured Query Language (SQL) processes in a task system are online tasks or batch tasks, leading to inefficient resource allocation and auto-scaling issues, particularly in cases involving local batch tasks, remote batch tasks, and web service batch tasks.

Innovation Solution

A control program that acquires processing request information and operation data from multiple servers to determine the type of processing request based on processing result information, first operation information, and second operation information, allowing for appropriate resource allocation and auto-scaling according to task characteristics.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If message-based determination techniques are used to classify tasks as online or batch, then task classification can be performed for messages exchanged within the system, but the technique fails to determine task type for local batch tasks executed by the DB server without communication

Engineering Contradiction:

Improvetask type determination accuracyVSAvoidcoverage of task types

Core Design Contradiction:

Measurement precisionVSAdaptability or versatility

Solution Approach 1:

The determination technique is segmented into multiple approaches: message-based determination for remote tasks, and process information-based determination for local tasks. This segmentation allows each method to specialize in specific task types, improving both accuracy and coverage across different task scenarios.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

Process information acts as an intermediary indicator to determine task types when message-based determination is not applicable. By using process information (process ID, start time, end time) as a mediator, the system can classify local batch tasks that do not involve message communication, thereby expanding the versatility of the determination technique.

Inventive Principle:

Principle #24Intermediary (Mediator)

2Reliability

If resources are allocated to maintain processing capacity for online tasks, then response time requirements are met, but resource utilization efficiency decreases during batch processing periods

Engineering Contradiction:

Improveresponse time guaranteeVSAvoidresource utilization efficiency

Core Design Contradiction:

ReliabilityVSProductivity

Solution Approach 1:

Resource allocation is made dynamic based on task type determination. The system adjusts resource allocation between online and batch tasks according to the determined task characteristics, allowing resources to be optimized for response time during online tasks and for processing efficiency during batch tasks, thereby improving both reliability and productivity.

Inventive Principle:

Principle #15Dynamics

Solution Approach 2:

The system changes allocation parameters (resource allocation, scaling decisions) based on the determined task type. By identifying whether a task is online or batch, the system can adjust parameters such as resource allocation and scaling behavior to meet the specific requirements of each task type, optimizing both response time guarantees and resource utilization.

Inventive Principle:

Principle #35Parameter changes

3Productivity

If auto-scaling is performed without accurate task type determination, then resource allocation may be inappropriate, leading to either over-provisioning or under-provisioning of resources

Engineering Contradiction:

Improveprocessing capacityVSAvoidunnecessary scaling costs

Core Design Contradiction:

ProductivityVSLoss of energy

Solution Approach 1:

The system implements feedback-based auto-scaling by using task type determination results to guide resource allocation decisions. The determined task type (online or batch) provides feedback that informs scaling decisions, ensuring that resources are allocated appropriately based on actual task characteristics rather than generic rules, thereby avoiding both over-provisioning and under-provisioning.

Inventive Principle:

Principle #23Feedback

Data Source

PatentUS10171620B2Non-transitory computer-readable recording medium having stored therein control program, control apparatus and control method

Publication Date: 2019.01.01 FUJITSU LTD

AI summary

Processing request information of a processing request for a first computer by a second computer is acquired, first operation information and second operation information that relate to operations of the first computer and the second computer are acquired, respectively; and a type of the processing request is determined based on processing result information, the processing request information, the first operation information, and the second operation information. The processing result information associates a processing history of the first computer according to the processing request included in the processing request information for the first computer with the second operation information that relates to the processing request by the second computer making the processing request. Accordingly, the type of the processing request can be appropriately determined.