RESTful Management Distributed Collection for Application Servers
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In application server environments, the existing methods for collecting and managing data from a large number of servers are inefficient, leading to slow response times and scalability issues, as they require direct RPCs to each runtime MBean, which burdens the client with complex logic and sequential or parallel requests.
Innovation Solution
Implementing a RESTful management system that allows managed servers to assist in the fan-out of REST searches across servers, reducing the burden on the admin server and improving scalability by grouping servers into nested batches for parallel processing.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the admin server directly fans out REST requests to each managed server, then the client receives comprehensive metrics data, but the response time becomes unacceptably long (25 seconds for 8,000 servers)
Solution Approach 1:
The patent divides the domain servers into nested batches, where each batch is processed independently. The admin server segments the fan-out operation into multiple levels: top-level batches processed by the admin server, and sub-batches processed by designated servers within each batch. This segmentation allows parallel processing of multiple batches simultaneously, reducing the overall response time from 25 seconds to approximately 0.15 seconds for 8,000 servers.
Solution Approach 2:
The patent implements a nested batch structure where batches contain sub-batches, which in turn may contain further sub-batches. Each batch is assigned to a designated server that independently processes its assigned sub-batches. This nested structure enables hierarchical parallel processing, where multiple levels of batches are processed concurrently, dramatically improving data collection efficiency while maintaining complete metrics coverage.
2Measurement precision
If the client directly requests metrics from each runtime MBean using RPC, then the client can obtain detailed server data, but the client becomes burdened with complex logic and sequential or parallel request management
Solution Approach 1:
The patent introduces designated servers as intermediaries between the client and the runtime MBeans. Instead of the client directly managing complex RPC requests to thousands of MBeans, the client simply sends a single REST request to the admin server. The admin server then distributes requests through designated servers in nested batches, which collect and consolidate data from their respective sub-batches. This intermediary structure eliminates client-side complexity while preserving detailed server data collection.
3Adaptability or versatility
If more servers are added to the domain to improve scalability, then the domain capacity increases, but the time to gather information grows proportionally
Solution Approach 1:
The patent implements a dynamic nested batch structure that automatically adapts to the number of servers in the domain. As servers are added or removed, the batch hierarchy is dynamically reconfigured to maintain optimal parallel processing efficiency. The system determines the number of top-level batches and sub-batches based on the total server count, ensuring that information gathering time remains constant regardless of domain size. This dynamic adaptation allows the domain capacity to scale from hundreds to thousands of servers without increasing response time.
Data Source
AI summary
In accordance with an embodiment, described herein is a system and method for RESTful management distributed collection in an application server environment. Distributed collection can be used to improve the performance of REST searches that fan-out across servers, by having the managed servers assist in the fan-out, instead of having the admin server fan-out to each managed server directly.