Serverless edge computing is gradually becoming the dominant force in application deployment within cloud computing. It allows developers to focus on building individual functions with application logic and pay only for the actual usage of these functions. Among them, interdependency functions are scheduled in a distributed manner within the edge network. However, the heterogeneity of edge networks and the complex dependencies of functions pose a huge challenge to achieving optimal scheduling decisions. In addition,it is also difficult to split data streams and map them to the paths between heterogeneous edge servers. To address this issue, we propose a method called Serverless Function Scheduling Optimization (SFSO). First, we design an innovative function merging module that incorporates both vertical and horizontal data merging techniques, which help avoid redundant data transmission overhead and ensure load balancing. Then, by solving several infinity norm minimization problems, we find the optimal split for each data flow and determine the best deployment scheme for each function. We compare SFSO with three contrasting methods. The experimental results show that, based on real serverless workflow data, SFSO significantly reduces the workflow running time, demonstrating the effectiveness and superiority of SFSO.