Together with the CPU and the network, storage plays an essential role in the overall performance of applications, especially in the context of big data applications, that can deal with enormous datasets. However, testbeds have scarce support for experiments involving storage performance. Experimenters can use the storage devices provided on the testbed directly, but (1) they might not provide the suitable performance characteristics for their experiments; (2) it might leave an uncontrolled bias over the experiments' results. In this paper, we explore the feasibility of using Linux's control groups to emulate I/O performance for testbed experiments. We then use it in the Distem emulator to create a customizable I/O experimental environment. Using Distem, we perform experiments on Hadoop to highlight the advantage of emulating I/O performance. Results obtained from a cluster of 25 nodes show how the performance of Hadoop changes according to emulated storage performance.