Optimizing Large Language Model Deployment in Edge Computing Environments