Copyright by William Lloyd Bircher 2010 - The Laboratory for ...
Copyright by William Lloyd Bircher 2010 - The Laboratory for ...
Copyright by William Lloyd Bircher 2010 - The Laboratory for ...
You also want an ePaper? Increase the reach of your titles
YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.
2.1.2 Subsystem-Level Power in a Server System<br />
To study component-level server power, the aggregate CPU power measurement<br />
framework is used and extended to provide additional functionality required <strong>for</strong><br />
subsystem-level study. <strong>The</strong> most significant difference between the studies of CPU level<br />
versus subsystem level is the requirement <strong>for</strong> simultaneously sampling multiple power<br />
domains. To meet this requirement the IBM x440 server is used which provides separate,<br />
measureable power rails <strong>for</strong> five major subsystems. It is described in Table 2.2.<br />
Table 2.2 Server System Description<br />
System Parameters<br />
Four Pentium 4 Xeon 2.0 GHz, 512KB L2 Cache, 2MB L3 Cache, 400 MHz FSB<br />
32MB DDR L4 Cache<br />
8 GB PC133 SDRAM Main Memory<br />
Two 32GB Adaptec Ultra160 10K SCSI Disks<br />
Fedora Core Linux, kernel 2.6.11<br />
By choosing this server, instrumentation is greatly simplified due to the presence of<br />
current sensing resistors on the major subsystem power domains. Five power domains<br />
are considered: CPU, chipset, memory, I/O, and disk. <strong>The</strong> components of each<br />
subsystem are listed in Table 2.3.<br />
Table 2.3 Subsystem Components<br />
Subsystem Components<br />
CPU Four Pentium 4 Xeons<br />
Chipset Memory Controllers and Processor Interface Chips<br />
Memory System Memory and L4 Cache<br />
I/O I/O Bus Chips, SCSI, NIC<br />
Disk Two 10K rpm 32G Disks<br />
15