Entropy-Based BDD Compression for Network Model Checking
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional network model checking using Binary Decision Diagrams (BDDs) faces the 'state explosion problem' due to exponential memory requirements, which is exacerbated by the difficulty in determining optimal variable ordering within an acceptable time frame, making it computationally hard to achieve efficient memory usage and computation time.
Innovation Solution
The proposed solution employs entropy-based BDD compression techniques, specifically ordering network fields based on calculated field-level entropy to reduce the size of BDDs, using bit-level reduction and field-level reduction methods to minimize memory usage and computation time, and applying domain-specific knowledge to optimize BDD performance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional BDD methods are used for network model checking, then the model checking can be performed, but the memory requirements grow exponentially due to the state explosion problem
Solution Approach 1:
The patent segments the network model into multiple components and represents each component's behavior as a separate BDD. By dividing the overall system model into smaller subsystem models, the exponential memory growth is mitigated while maintaining model checking capability. This is achieved by creating BDDs for individual network entities rather than the entire network at once.
Solution Approach 2:
The patent implements a hierarchical BDD structure where BDDs for network entities are nested within a global BDD representation. The global BDD contains references to entity-specific BDDs, allowing the system to manage large network models by nesting smaller BDD structures within a larger framework, thereby reducing overall memory requirements.
2Quantity of substance
If optimal variable ordering is pursued to reduce BDD size, then memory usage improves, but the computation time to determine optimal ordering becomes excessively long
Solution Approach 1:
The patent performs preliminary variable ordering for each network entity's BDD before constructing the global BDD. By pre-ordering variables at the entity level using heuristics, the system avoids the computationally expensive task of optimizing the entire global BDD variable ordering, thus reducing computation time while achieving acceptable BDD size.
Solution Approach 2:
The patent employs dynamic variable ordering strategies where the ordering is adapted based on the specific characteristics of each network entity and the current state of BDD construction. Rather than using a fixed ordering scheme, the system dynamically adjusts variable ordering during the BDD construction process to balance memory efficiency with computation time.
3Measurement precision
If detailed network models are created for accurate verification, then verification precision improves, but the preprocessing time and computation resources increase significantly
Solution Approach 1:
The patent segments the verification process into entity-level verification and global verification phases. By performing verification on individual network entities first using their own BDDs, the system achieves precise verification for each component without requiring the entire network model to be fully constructed and processed, thus reducing preprocessing time while maintaining verification accuracy.
Solution Approach 2:
The patent implements partial verification by focusing on critical network entities and properties that most impact overall system correctness. Rather than performing exhaustive verification on all possible network behaviors, the system selectively verifies key aspects, achieving sufficient precision for practical purposes while significantly reducing computation time.
Data Source
AI summary
Methods, systems and media for network model checking using entropy based binary decision diagram (BDD) compression are described. Two related compression techniques are described: bit level reduction to reduce the number of bits required for each network field according to its nature, and field level reduction to reduce the size of the BDD tree by finding a near-optimum ordering of the fields in the BDD space. These two techniques, separately or together, may alleviate the state explosion problem the limits application of BDD based model checking. The two techniques complement each other synergistically, particularly in the domain of computer network checking and verification.