Cloud Request Identification Through Semantic Event Clustering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing cloud computing technologies struggle to identify and optimize relevant requests within workloads due to unclear event triggers and challenges in determining the correct sequential arrangement of events, particularly in distributed processing environments, leading to inefficient resource utilization and increased costs.
Innovation Solution
A method utilizing machine learning models to generate vector embeddings for raw data events, cluster these events into relevant and irrelevant categories, and identify requests as sequences of events, employing a two-stage training process to enhance accuracy and efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If traditional cloud computing resource allocation methods are used, then resource utilization can be improved through basic optimization techniques, but the core problem of identifying relevant requests in distributed processing environments remains unsolved
Solution Approach 1:
The patent replaces traditional mechanical/event-based request detection mechanisms with machine learning-based semantic understanding. Instead of relying on explicit event triggers or sequential patterns, the system uses transformer models to generate vector embeddings of code and events, enabling semantic similarity matching to identify relevant requests in distributed environments.
Solution Approach 2:
The patent changes the fundamental parameters for request identification from explicit event markers or sequential patterns to semantic vector representations. By transforming code and events into vector embeddings and using cosine similarity as the matching criterion, the system achieves robust request identification in distributed processing environments where traditional event-based approaches fail.
2Measurement precision
If machine learning models are used to generate vector embeddings and cluster events, then request identification accuracy is improved, but computational complexity and training requirements increase
Solution Approach 1:
The patent applies preliminary action by pre-training transformer models on large code corpora before actual request identification tasks. The models generate vector embeddings in advance, and clustering algorithms pre-organize event patterns, enabling fast and accurate request identification during runtime without requiring complex real-time computation.
Solution Approach 2:
The patent uses copying by creating vector embeddings as simplified representations of complex code and event structures. Instead of working with the full complexity of source code and event logs, the system copies essential semantic information into compact vector representations, reducing computational complexity while maintaining identification accuracy.
3Reliability
If all events are analyzed without filtering, then comprehensive workload analysis is achieved, but computational resources are wasted on irrelevant events
Solution Approach 1:
The patent extracts relevant events from the complete event log by using machine learning models to identify and select only the events that are semantically similar to the target request. The transformer models generate vector embeddings that capture essential event characteristics, enabling the system to extract and analyze only relevant events while filtering out unnecessary ones, thus reducing computational energy consumption.
Solution Approach 2:
The patent applies local quality by tailoring the analysis to specific local regions of interest in the event stream. Instead of uniformly processing all events, the system uses semantic similarity metrics to identify and focus on local event clusters that are relevant to the request, applying computational resources only where needed to maintain both completeness and efficiency.
Data Source
AI summary
A system and method for identifying a request of a service. The method includes generating vector embeddings for raw data of events using a machine learning model, wherein the machine learning model is trained to indicate semantic meaning of at least one event of the raw data of events; clustering, based on the vector embeddings, the at least one event of the raw data of events into a plurality of clusters, wherein a subset of the plurality of clusters includes relevant events in the data of events; and identifying a request as a sequence of events from the subset of the plurality of clusters.