To effectively address the service gap in the first and last mile connections of metro systems, a two-stage model was developed for designing feeder bus services. The first-stage was the feeder bus stop selection model, which was used to select feeder bus stops from existing conventional bus stops. In this stage, we proposed a method to merge stops to prevent feeder bus stops from being too close to each other. The second-stage was the feeder bus route planning model, which was used to plan routes between the feeder bus stops selected in the first-stage. A two-stage heuristic algorithm was established to efficiently solve the two-stage model. The practicality of the proposed approach was verified in Harbin. The results showed that: 1) bus stop merging method could reduce the stop overlap rate by 59.32% while decreasing the stop coverage rate by only 4.07%; 2) stop selection impacts route planning, and their integration in the two-stage model could ensure optimized feeder bus services; 3) compared with the conventional model, the two-stage model could reduce the total travel time by 27.17%; and 4) appropriately increasing the number of routes could reduce the average route length. The proposed method provides a reference for designing feeder bus services.

Containerization has garnered substantial favor among cloud service providers. Nevertheless, the notable network overhead incurred between containers has prompted concerns within the community. In cloud resource scheduling, collocating service containers that frequently communicate to the same machine - termed “service affinity” - is instrumental in enhancing application performance. In response to this concern, we present a solution that harnesses service affinity and collocates containers to enhance the overall system performance and stability. To maximize the benefits of collocating containers, it is necessary to calculate a new schedule that optimally and efficiently maximizes service affinity, especially within the expansive domain of industry-scale cloud environments. In pursuit of this, we leverage the skewness property of affinity and machine learning to fuse solver-based algorithms, thereby assuring both quality and efficiency for problems at scale. Our methodology encompasses the partitioning of a given task into discrete subproblems, with a keen focus on resolving the most critical ones. Via a graph neural network classifier, we assign each subproblem to be solved independently using methods based on off-the-shelf solvers in our algorithm pool - namely, MIP-based, or column generation. This strategic approach enables the efficient computation of a schedule for a cloud cluster that fully optimizes the overall service affinity. We further propose a heuristic algorithm to compute executable container migration plans for practical use, facilitating the transition to the new placement where service affinity is well optimized. Our solution has been deployed in our large-scale production environment, covering over a million cores within ByteDance. Through the successful real-world production deployment, our approach exhibits an average improvement in end-to-end latency by 23.75% and a reduction in request error rates by 24.09% compared to the original system.