Consider a scenario where you are working with Azure Synapse Analytics to process large-scale analytical workloads. The data includes sensor readings from multiple devices, each recording different metrics at varying intervals. What partition strategy would you implement to ensure optimal performance and scalability?