In our various applications at start-up and periodically we check if
there is an internet connection available. If so, and a known cloud has
been identified, the application will check for usability and then ask a
user if calculations are to be performed using cloud resources. The
application may desire to upload or download database updates as
well. These would normally include any updates or corrections to
history - perhaps Mozambique finally sends along a count of doctors
working in 2007. Recalling from the raw data Mozambique had 514
doctors working in 2004 and 548 in 2006. No data reported for 2005,
2007, 2008 and 2009. Perhaps the newly provided value for 2007 was
574 and we had previously filled in 531 in 2005, 564 in 2007, 579 in
2008 and 593 in 2009. Since Mozambique is a big place (about 800,000
square kilometers - roughly twice the area of California) , and, despite
AIDS, the home of more than 22 million people (about 3/5 the
population of California), it is likely we would have used a range of
values for doctors working. In point of fact, we used values between 282
and 1128. Sadly, Mozambique is terribly short of medical professionals
and not much positive happens even if the number of doctors
miraculously doubles. Even one more doctor would matter to the
hundreds or thousands of people he or she treats.
We have a modest sized vector of measures by country for 2010. What
we would like to know is are 40-odd measures present for all 200+
countries?
We'll assume we were able to construct two vectors that translate
measures and countries to integers. With that foundation, we will build
a three-dimensional (year - 2000, equivalent of country, equivalent of
measure) array in memory to reduce database contention and allow
parallelism. Now we have for each new value three questions: (1) how
does this compare to past values for this country and measure (2) how
does the number of doctors working in Mozambique in 2010 compare to
the number of doctors working in ostensibly similar countries and (3)
how does the number of doctors compare to changes in other 2010
measures for Mozambique?
Once it appears all data is present, before we begin to verify values
using covariance, we first determine what computing resources are
available.