Calculating the most reliable maximum flow (MRMF) from the edge cache node to the requesting node can provide an important reference for selecting the best edge cache node in a content delivery network (CDN). However, SDBA, as the current state-of-the-art MRMF algorithm, is too complex to meet real-time computing needs. This paper proposes a set of MRMF algorithms: NWCD (Negative Weight Community Deletion), SCPDAT (Single-Cycle Preference Deletion Approximation algorithm with Time constraint) and SCPDAP (Single-Cycle Preference Deletion Approximation algorithm with Probability constraint). NWCD draws on the “flow-shifting” algorithm of minimum cost and maximum flow, and further defines the concept of negative weight community. This algorithm continuously deletes the negative weight communities, which can increase reliability while keeping the flow constant in the residual graph. It is proven that when all negative weight communities are deleted, the corresponding maximum flow is the MRMF. SCPDAT tries to approach the optimal solution to the greatest extent possible within the limited time, while SCPDAP tries to reach the probability threshold in the shortest amount of time. Both of these adopt the strategy of first deleting single-cycle communities (which contribute more to the reliability with lower time cost). Experiments show that, compared with SDBA, NWCD combined with the probabilistic pruning achieves an order of magnitude improvement in time cost, while SCPDAT and SCPDAP demonstrate better time performance and increased applicability.
Bibliographical noteFunding Information:
Funding Statement: This work was partly supported by Open Research Fund from State Key Laboratory of Smart Grid Protection and Control, China (Zhang B, www.byqsc.net/com/nrjt/), Rapid Support Project (61406190120, Zhang B), the Fundamental Research Funds for the Central Universities (2242021k10011, Zhang B, www.seu.edu.cn) and the National Key R&D Program of China (2018YFC0830200, Zhang B, www.most.gov.cn).
- Content delivery network
- Flow reliability
- Maximum flow
- Uncertain graph