IBM 88743BU - System x3950 E, User Manual

Page: 214 / 426

196
Planning, Installing, and Managing the IBM System x3950 M2
4.1 Introduction and terminology
A multinode configuration is multiple individual servers connected together by a
high speed bus to form a single system image. This system is seen in an
installed operating system as one system and the operating system has access
to all resources in each server.
We often refer to this multinode configuration as a
complex
or an
n-node scalable
system
. Each individual server within this complex is called a
node
.
The following terms are used when describing a multinode configuration:
򐂰 Node
An x3950 M2 server that contains at least one processor, one memory card
with a pair of DIMMs installed, and has the capability to scale.
򐂰 1-node system or stand-alone system
This is a server that has been temporarily removed from a complex, or has
been booted separately for maintenance.
򐂰 2-node, 3-node, 4-node systems
A scaled system of two nodes, three nodes, or four nodes.
򐂰 Complex
This is at least two systems that have the ScaleXpander chip installed as
described in 4.5, “Upgrading an x3850 M2 to an x3950 M2” on page 204, and
are cabled with the scalability cables as described in 4.6, “Cabling of
multinode configurations” on page 209.
򐂰 Complex Descriptor
This is a data record that is generated by a BMC in an x3950 M2 and that is
capable in scaling of partitions. A ScaleXpander chip is required and installed.
The data record contains local system-specific information, and the system
state and partition information of all other systems in this complex.
See details in 4.3.2, “Complex Descriptor contents” on page 200.
򐂰 Merging
This is when the nodes are in the process of forming a multinode complex to
bind their resources to one system.
Progress messages appear in the POST, followed by a Merge completed
message. During this process you have the option to cancel the merge
process and boot each node individually in an
unmerged state (such as for
maintenance). See “Start a multinode system” on page 234.