As suggested by Masoud, GAMS is certainly the most appropriate software to implement DEA as you will have to write explicitly the linear programming and understand the mathematical structure of it.
Nonetheless, two main drawbacks of GAMS are its non-friendly user interface and its price. These reasons explains why plug-and-play software are now available online (free or not), and allow implementation of the method at a low programming-time cost.
R software and many packages developed lately (benchmarking, fear...) are also a good opportunity to implement DEA without entering into the programming part, but you should consider your result carefully as you don't have a fully clear vision on the underlying mathematical program.
To be synthetic, you can implement DEA in few hours on plug-and play software or function, or in many monthes or years on linear programming software, depending on how in-depth you want to understand the method, its mathematical and economical assumptions, and their consequences on the structure of the model.
R and softwares online can maybe be a good eye-opener on DEA potential, but if you pursue in using the method, i would recommend as Palani to invest time (and money) in GAMS, which will give you a very generic platform to implement any kind of model, and their statistical robustness test.