Edge computing enables a wide variety of application services for the Internet of Things, including those with performance-critical requirements. To achieve this, it brings cloud computing capabilities to network edges. A key challenge therein is to decide where and when to place or migrate application services considering their load variation and seeking the optimization of multiple performance objectives. In this paper, we address this optimal service placement issue by further considering how to distribute the load of an application placed in different locations. By estimating the performance-cost trade-off of services migration, we propose a dynamic service placement and load distribution strategy that uses limited look-ahead prediction to handle load fluctuations. Evaluation analysis demonstrates that our proposal outperforms other benchmarks solutions in terms of multiple conflicting objectives.