Best Practices When Deploying Linux on the HP ProLiant DL980 (updated December 2012)

7
Figure 3: Enabling OS Control Mode using RBSU
Non-Uniform Memory Access (NUMA) overview
The latest Intel® Xeon® processors provide Socket Local Memory, which consists of embedded memory controllers that
access memory DIMMS connected to the socket. A processor and the associated Socket Local Memory are considered a
node. Each node can be thought of as a separate Symmetric Multi-Processing (SMP) system where accesses to the
Socket Local Memory have the same latency. However, memory accesses from a processor to the physical memory (or
cached memory) of another node have greater latencies. The resulting architecture of a system having different
latencies for accessing memory is known as a NUMA architecture.
Figure 4 illustrates a simplified view of the DL980 architecture. Each pair of processor sockets, such as 0 and 1, has a
dedicated Quick Path Interconnect (QPI) link between them and comprises a QPI island. The interconnect node
controllers (XNC) interconnect the QPI islands.