IBM Power 770, Технический обзор и введение

Страница: 175 / 220

Chapter 4. Continuous availability and manageability
161
4.1 Reliability
Highly reliable systems are built with highly reliable components. On IBM POWER
processor-based systems, this basic principle is expanded upon with a clear design for
reliability architecture and methodology. A concentrated, systematic, architecture-based
approach is designed to improve overall system reliability with each successive generation of
system offerings.
4.1.1 Designed for reliability
Systems designed with fewer components and interconnects have fewer opportunities to fail.
Simple design choices, such as integrating processor cores on a single POWER chip, can
dramatically reduce the opportunity for system failures. In this case, an 8-core server can
include one-fourth as many processor chips (and chip socket interfaces) as with a double
CPU-per-processor design. Not only does this case reduce the total number of system
components, it reduces the total amount of heat generated in the design, resulting in an
additional reduction in required power and cooling components. POWER7 and
processor-based servers also integrate L3 cache into the processor chip for a higher
integration of parts.
Parts selection also plays a critical role in overall system reliability. IBM uses three grades of
components (1, 3, and 5), with grade 3 defined as the industry standard (“off-the-shelf”). As
shown in Figure 4-1, using stringent design criteria and an extensive testing program, the IBM
manufacturing team can produce grade 1 components that are expected to be 10 times more
reliable than industry standard. Engineers select grade 1 parts for the most critical system
components. Newly introduced organic packaging technologies, rated grade 5, can achieve
the same reliability as grade 1 parts.
Figure 4-1 Component failure rates
Component failure rates
0
0.2
0.4
0.6
0.8
1
Grade 3 Grade 1 Grade 5