Hitachi Compute Blade 2000 User Manual page 1073

When Scheduling mode is shared mode
In the case of the Shared mode, the memory access of logical processors are generally
dispersed to multiple nodes, and access to local memory and remote memory are
intermingled.
The memory access performance of all LPARs in shared mode will be between the local
one and remote one, and there will be almost no difference between LPARs. (See
LPAR 1 and LPAR 2 below.)
LPAR1
1way
LPAR2
2way
Physical processor
Socket
Core
small
Physical memory
Node0
When the processor group is not configured,
The memory access performance of all the LPAR in share mode will be between local one and
remote one.
Memory access of processors in dedicated mode
without any processor group
Logical processor
Time-shared
small large large
Node1
1022
Memory access
Memory
access cost
Access to the remote
memory and access to
the local memory are
intermingled.
*Shows only the memory access of
LPAR2.
(LPAR1 would be the same. )