IBM RS/6000 SP, Руководство по определению проблем

Страница: 22 / 316

This soft copy for use by IBM employees only.
Figure 1. Control Workstation Connected to the SP Frame
The serial link connects the CW with the Supervisor Card into the Frame. In this
way, the CW can manage and monitor every major hardware event produced
either by the Frame itself or by the nodes. Each node has a Node Supervisor
Card which is connected to the Frame Supervisor Card, as shown in Figure 2 on
page 3. The new Supervisor Card is able to connect a second CW as a backup
Control Workstation by using the HACWS software option for the POWERparallel
System Support Programs.
Note: More information about HACWS can be found in High Availability on SP,
SG24-4742.
The Ethernet network (called SP Ethernet) connects each node and the CW.
Usually this network is dedicated to SP traffic only, therefore you should install a
different network adapter to pass user applications traffic.
1.1.2 The Frame
All the hardware information flows from the Frame to the CW through the Serial
Link. In this way, the CW is able to detect hardware failures in the nodes or in
the Frame, and take actions over hardware components.
However, the CW uses the Ethernet to recognize when the nodes are “alive,” not
just powered on, but up and running. Communications with the nodes is handled
by the hardware daemon (called hardmon), which sends small packages called
heartbeat packages through the TCP/IP stack to get in contact with the nodes.
When the nodes are up and running, the hardware monitor sets the
host
responds variable from the SDR to “1” for each node.
At this time, the hardware monitor can use only the SP Ethernet to send
heartbeat packages, which reduces this network to a single point of failure. But
2
SP PD Guide