Placement Of Components - IBM Power System E850C Technical Overview And Introduction

IBM uses stringent design criteria to select server grade components that are extensively
tested and qualified to meet and exceed a minimum design life of seven years. By selecting
higher reliability grade components, the frequency of all failures is lowered, and wear-out is
not expected within the operating system life. Component failure rates can be further
improved by burning in select components or running the system before shipping it to the
client. This period of high stress removes the weaker components with higher failure rates,
that is, it cuts off the front end of the traditional failure rate bathtub curve (see Figure 4-1).
Figure 4-1 Failure rate bathtub curve

4.2.2 Placement of components

Packaging is designed to deliver both high performance and high reliability. For example, the
reliability of electronic components is directly related to their thermal environment. Large
decreases in component reliability are directly correlated to relatively small increases in
temperature. All POWER processor-based systems are packaged to ensure adequate
cooling. Critical system components, such as the POWER8 processor chips, are positioned
on the system board so that they receive clear air flow during operation. POWER8 systems
use a premium fan with an extended life to further reduce overall system failure rate and
provide adequate cooling for the critical system components.
The Power E850C has two cooling channels. The front fans provide cooling for the upper part
of the chassis, covering the memory cards, processors, and PCIe cards. These fans in this
assembly provide redundancy, and support concurrent maintenance. The lower system fans,
which are in the internal fan assembly, provide air movement for the lower part of the chassis,
including the disk backplane and RAID controllers. The fans in the internal fan assembly
provide redundancy, and also contain multiple integrated spares.
Wafer and chip test
In situ burn-in
System
run-in
Early life
Power-on Hours
Average life
of a server
Chapter 4. Reliability, availability, and serviceability
End-of-life
105