When it comes to upgrading the world's
largest payment processing system, failure is not an option, said Scott
Thompson, executive vice president of Visa International's technology practice,
Inovant LLC. Thompson recently shared some of his best practices with eWEEK
Labs.
Teamwork. Two years ago, Visa
determined that completed projects would be more efficient and more flexible if
quality assurance and engineering staff work closely together on the development
of code.
Visa made a strategic decision to make engineering
divisions responsible for quality assurance. This way, everyone involved
with the project is equally vested in its outcome, Thompson said. This is
especially important because the amount of time it takes to write an October release
has jumped from 40,000 worker hours three years ago to more than 150,000 worker
hours this year with no increase in personnel.
Responsibility. One of the main reasons
Visa has been able to pull off its network upgrades is that everyone involved on
the projects knows exactly how their decisions will affect the business. During
the planning process, business managers and engineers discuss the types of
changes they'd like to see and then determine what needs to be changed within
the code to make those changes happen.
"The fact that everyone knows exactly
how these changes are going to impact Visa as a business is paramount to our
success," said Robin Owens, division head for Visa's Member Information Clearing
and Settlement.
Work closely with partners. The company
can't rely on its 21,000 merchant and bank partners to make changes as quickly
as it does. As a result, Visa often throttles back its project to ensure that
all its partners are on the same page and can make the upgrade jump
together.
Visa also works closely with partners
to help them prepare for the network updates.
Stick to schedule. At any point during
the implementation, Visa engineers know exactly where they are in the upgrade
cycle. Before the upgrade begins, a script is distributed to employees
throughout the organization so that everyone knows at exactly what time a
particular data center will be upgraded and exactly how long that upgrade will
take.
"We don't do something and hope it's
going to take 2 hours; we know it's going to take 2 hours," Thompson said. "We
know exactly what's going to happen."