The attention to "Big Data" and "Clouds" sometimes hides the importance of an (old?) core topic of computer science: programming.  Big data sets amplify any system or software inefficiencies when large data sets are processed, stored, or transmitted. So the efficiency of programs matters (again) if we want to work with applications based on big data or in a cloud.  In this talk I will talk about (i) how we can prepare computer scientists for this environment (teach programming of parallel systems early in the curriculum) and (ii) what resources we need to teach the students (in our case , the "Datacenter Observatory", an platform for system level teaching at research supported by EPFL and ETH Zurich).