In the New Radio Access Network architecture (New RAN), currently envisioned by 3GPP, the evolved NodeB (eNB) functions can be split between a Distributed Unit (DU) and Central Unit (CU). Furthermore, as per the Virtual RAN (VRAN) approach, such functions can be virtualised (e.g., in simple terms, deployed in virtual machines). In such scenario, the fronthaul network connecting DU and CU must fulfill different latency and capacity requirements based on the selected functional split. This study experimentally evaluates in a federated testbed how the fronthaul latency budget (i.e., the latency requirement of the fronthaul network connecting DU and CU), specified by Standard Developing Organisations (SDO) (3GPP in this specific case), is impacted by virtualising some of the RAN functions. In particular, Option 7-1 (i.e., intra-PHY split) and different virtualisation methods are considered for the CU. Furthermore, it evaluates how jitter (i.e., delay variation) impacts the DU-CU communication. The obtained results show that light virtualisation methods (e.g., Docker) impact less the fronthaul latency budget than heavy virtualisation methods (e.g., VirtualBox). In addition, a maximum jitter of about 40 μs can be tolerated in the fronthaul. © 2017 IEEE.