When should you use SVM?

Could you elaborate on the scenarios where Support Vector Machines (SVM) would be the most suitable choice for a machine learning task? Are there specific types of data or problems that SVM tends to excel at solving? Additionally, what are some of the potential drawbacks or limitations of using SVM that practitioners should be aware of before implementing it in their projects?