Joshua Resnick - Fotolia
There is tremendous interest in moving applications and data to the cloud these days. But, for existing applications, there are a lot of decisions to make and steps to take to get there.
Planning such moves was the topic of discussion at a recent conference where an analytics manager at a leading merchant payments services company shared some snapshots from his company's journey to a cloud payment processing system.
The message was that cloud can simplify important infrastructure deployment tasks that take up IT's time, but that making the move to the cloud requires careful consideration -- and trying things out.
"You have to think about things differently on the cloud. We iterated in order to figure it out," said Mark Kubik, vice president of BI and analytics at Global Payments Inc. Kubik spoke at last month's Data Summit 2018 in Boston.
In his case, iterated meant picking a portion of a larger payment tokenization application for migration, and then getting early feedback from customers to make sure things were on the right track.
The first application parts tapped for the move to the cloud were elements of a portal for merchant customers of Global Payments services. This required creating a data lake on the cloud for merchant, transaction and other data.
As described at the conference, an early element of the cloud migration was a chargeback system to handle payment refunds.
"One of our first drivers from the beginning was getting customer feedback. We put the user interfaces out there so they could test it," Kubik said, admitting that the first iteration "looked like IT guys had designed it."
Mark Kubik VP of BI and analytics, Global Payments
Kubik said Atlanta-based Global Payments is pursuing a multi-cloud strategy, with "different workloads for different clouds."
For its earliest cloud payment processing system efforts, Global Payments is using the Google Cloud Platform, in large part because of the capabilities of that cloud's BigQuery managed service for data warehousing and analytics, according to Kubik. He said Google, as well as tools and consulting services company Bitwise, were like partners to Global Payments in the initial deployment project.
Bitwise helped Global Payments with migration planning; automated conversion of legacy extract, transform and load (ETL) code; and populating the Google Cloud data lake, said Shahab Kamal, executive vice president for solutions engineering at Bitwise. He said data conversion is often one of the factors that make cloud migration difficult.
"For large companies, the timeline and cost of ETL conversion is so long and large that they don't want to pursue it," Kamal said.
Data infrastructure on the cloud
Both Kamal and Kubik said Google Cloud's capabilities in AI were seen as beneficial in future stages of the cloud payment processing effort, which would see machine learning and other advanced analytics techniques applied for the purposes of Global Payments' merchant customers.
"Machine learning comes built in with the Google Cloud," Kamal said.
The complexities of fielding infrastructure for big data analytics were a factor in moving to the cloud, Kubik indicated.
"If it takes a long time to stand up infrastructure, I can't get products out quickly. [On premises,] we had to script everything," Kubik said.
For comparison, he said, Global Payments was able to put the entire infrastructure for a development environment on the Google Cloud in 15 minutes. He said test environments have moved over, too, and the first production application is now running on the cloud.
Kubik said early adopters within the organization who used spreadsheets to access reports from the cloud payment processing system saw some limits on how many rows of data they could download. However, that issue has since been addressed.
Providing access to multiple BI tools is a key part of the plan. "We allow choices," Kubik said.
How good have cloud vendors been in responding to Global Payments' needs?
"The cloud vendors seem receptive to help us," he said. "They are all in a race."