Although the problem of QoS-aware resource allocation is not new, novel hardware-based resource allocation mechanisms have recently become available in commodity cloud servers and enabled a new generation of QoS-aware resource allocation approaches. Unfortunately, to the best of our knowledge, existing proposals are by design tailored to single-socket architectures only. In many warehouse scale data centers, dual-socket (or even larger) machines already constitute the largest share of hosts. This paper presents the full design and implementation of BALM, a QoS-aware memory bandwidth allocation technique for multi-socket architectures. BALM combines commodity bandwidth allocation mechanisms originally designed for single-socket with a novel adaptive cross-socket page migration scheme. Our evaluation with a large and dynamic set of real applications co-located on a dual-socket machine shows that BALM can overcome the efficiency limitations of state-of-the-art. BALM delivers substantial throughput gains to bandwidth-intensive best-effort applications, while ensuring marginal SID violation windows to latency-critical applications.