Sliding Window Cache for Network Telemetry Query Latency

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Current data observability techniques face performance issues due to the large volume and frequent updates of operational data, leading to query latency and resource consumption, especially when supporting queries with subscription and continuous queries for operational data in network management systems.

Innovation Solution

Separating operational data queries from configuration data queries and employing a sliding window cache to process and store operational data, allowing for efficient retrieval and notification of changes while reducing computational resources needed for continuous queries.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If operational data is queried with filtering and ordering operations, then query results are obtained, but query latency increases and system performance degrades

Engineering Contradiction:

Improvequery result accuracyVSAvoidquery latency

Core Design Contradiction:

Measurement precisionVSLoss of time

Solution Approach 1:

The patent pre-computes and stores filtered and ordered results in materialized views before queries are executed. When a query is made, the system retrieves pre-prepared results rather than performing expensive filtering and sorting operations on raw operational data, significantly reducing query latency while maintaining result accuracy

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The patent introduces materialized views as intermediary structures between raw operational data and query requests. These materialized views act as pre-processed intermediate results that satisfy query requirements, eliminating the need to perform expensive filtering and ordering operations on the original large-volume operational data

Inventive Principle:

Principle #24Intermediary (Mediator)

2Extent of automation

If subscription queries are supported for operational data, then real-time notifications are provided, but computing resources are significantly consumed

Engineering Contradiction:

Improvereal-time notification capabilityVSAvoidcomputing resource consumption

Core Design Contradiction:

Extent of automationVSUse of energy by moving object

Solution Approach 1:

The patent pre-evaluates subscription queries against materialized views so that when data changes occur, the system only needs to check if the pre-computed materialized views have changed rather than re-evaluating the entire subscription query. This dramatically reduces the computing resources needed for real-time notifications while maintaining automated change detection

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The patent implements a feedback mechanism where materialized views are updated incrementally based on change detection. When operational data changes, the system evaluates whether the change affects existing materialized views and only re-computes what is necessary, optimizing resource usage while maintaining real-time notification capability

Inventive Principle:

Principle #23Feedback

3Ease of operation

If large volume operational data is brought into memory for processing, then query operations can be performed, but system memory usage increases and performance decreases

Engineering Contradiction:

Improvequery operation capabilityVSAvoidmemory usage

Core Design Contradiction:

Ease of operationVSQuantity of substance

Solution Approach 1:

The patent extracts only the necessary pre-computed results into materialized views that are stored separately from the main operational data. This allows query operations to be performed on compact materialized views rather than requiring the entire large-volume operational data to be loaded into memory, significantly reducing memory usage while maintaining query capability

Inventive Principle:

Principle #2Taking out (Extraction)

Solution Approach 2:

The patent creates materialized views as copies of query results derived from operational data. These copies contain only the processed information needed for queries, eliminating the need to load and process the full volume of original operational data in memory, thus reducing memory consumption while preserving query functionality

Inventive Principle:

Principle #26Copying

Data Source

PatentUS20250016076A1Configuration and operational network observability using sliding window for append-only telemetry log

Publication Date: 2025.01.09 CISCO TECHNOLOGY INC

AI summary

Techniques are described for storing and processing network data for responding to queries for such network data. Operational network data is separated from configuration network data so that they can be processed and stored separately. A sliding window cache is used to continually, temporarily store network data objects having time stamps falling within the time range of the sliding window cache. Network data objects stored within the sliding window cache are then moved to computer memory for storage and later retrieval. In response to a query for network data, network data objects can be retrieved from the sliding window cache and also from the computer memory based on time stamps of the network data objects and on the time range of the query.