> does naybody know if Google's qualified Planners are looking at the
> TCO (total cost of ownership) to justify 100k servers versus
> SUN/AIX/zOS/zVM solutions?
> or is it philosophical and their investors/board of directors want to
> show innovation as a way to increase capitalization?

They watch this very closely. Their 100K servers are directly derived
from: 

1) The Google algorithm(s) are computationally intensive as well as I/O
intensive (more MIPS=better search performance).

2) The workload is rarely sequential processing (ie every new piece of
work is essentially a new problem, so prefetch or setup doesn't really
help much)

3) They need very low price/MIP ratios. 1U netbooted boxes (most of
their machines don't even HAVE disks, just CPU and network) are
essentially disposable at the volume that Google buys them -- 5000 to
10000 at a time. It is *literally* cheaper to do machine-level
replacement -- repair costs more than just dumping it and replacing it
with a newer/faster machine. It also provides them with a rolling
technology upgrade capability without shutdown for maintenance.

AOL has a similar strategy -- don't repair anything, replace it. 

----------------------------------------------------------------------
For LINUX-390 subscribe / signoff / archive access instructions,
send email to [EMAIL PROTECTED] with the message: INFO LINUX-390 or visit
http://www.marist.edu/htbin/wlvindex?LINUX-390

Reply via email to