Warehouse-Scale Computers to Exploit Request-Level and Data-Level Parallelism - Computer Architecture: A Quantitative Approach - page 439

Figure 6.4 shows the original weightings and the new weighting that match Fig-
ure 6.3 . These weightings reduce the performance summary by 30% from 3210
ssj_ops/wat to 2454.
FIGURE 6.4 SPECPower result from Figure 6.17 using the weightings
from Figure 6.3 instead of even weightings .
Given the scale, software must handle failures, which means there is litle reason to buy
“gold-plated” hardware that reduces the frequency of failures. The primary impact would be
to increase cost. Barroso and Hölzle [2009] found a factor of 20 difference in price-performance
between a high-end HP shared-memory multiprocessor and a commodity HP server when
running the TPC-C database benchmark. Unsurprisingly, Google buys low-end commodity
servers.
Such WSC services also tend to develop their own software rather than buy third-party
commercial software, in part to cope with the huge scale and in part to save money. For ex-
ample, even on the best price-performance platform for TPC-C in 2011, including the cost of
the Oracle database and Windows operating system doubles the cost of the Dell Poweredge
710 server. In contrast, Google runs Bigtable and the Linux operating system on its servers, for
which it pays no licensing fees.
Given this review of the applications and systems software of a WSC, we are ready to look
at the computer architecture of a WSC.