Unclouding: How one company reversed the cloud migration process
As cloud use matures, more companies are considering scaling back and starting a reverse cloud migration process. Here is one IT officer's 'unclouding' experience.
When Ijaaz Ullah was vice president of IT and privacy officer for Nightingale Informatix Corp., the largest cloud-based electronic medical records provider in Canada, he was responsible for two systems: One system, Nightingale On Demand or NOD, ran on a private cloud and another, Nightingale v10, was being tested to run in a public cloud.
The public cloud system, built out in the company's data center, was running in beta at 110 clinics in North America when the roadmap for its official rollout changed dramatically.
In the fall of 2016 -- just a few months after v10's cloud debut in June -- Nightingale Informatix was acquired by TELUS, the Canadian telecom company. TELUS decided not to take v10 to market. Development of v10 continued on premises for three months while the acquisition progressed, because the company still thought the product could be sold, Ullah said.
"I was disappointed to hear that the v10 software was not going to be taken to market, as we had spent a great deal of time in building a highly automated system that we were very proud of," said Ullah, now director of technical strategy and delivery at TELUS Health. "With that said, what we've learned from building and deploying this product has been valuable and this knowledge is transferrable."
Ullah's team had spent three years developing v10, building out the cloud infrastructure and associated integrations to do automation. The public cloud beta version had been trotted out at tradeshows and was running at all those clinics -- but the path forward was clear. When the acquisition of Nightingale by TELUS closed, all the work on launching a public cloud version of the medical records system would cease and Ullah and his team would get to work at "unclouding" v10.
The 'unclouding' trend
Ullah and his team are not the only ones unclouding workloads.
As companies have become more knowledgeable about the costs and benefits of cloud computing versus keeping workloads on premises, some are hitting the reverse button. Almost 40% of organizations with public cloud experience have migrated systems back from the public cloud to in-house data centers, according to a recent Datalink/IDG survey of over 100 IT professionals. The top reasons cited for the move were security (55%), cost/pricing concerns (52%) and manageability (45%).
Companies may discover that they desire stronger security on a given application than the cloud provider can offer, and may choose to move back on premises to gain a greater degree of control.
Seth Robinsonsenior director of technology analysis, CompTIA
A September 2016 study by CompTIA of 500 business and IT executives found almost identical reasons for the decision to move systems back on premises: security (58%), failure to achieve cost goals (30%), failure to integrate (24%) and dissatisfaction with reliability (22%). The issue of security, however, is paramount, if not permanent.
"Companies may discover that they desire stronger security on a given application than the cloud provider can offer, and may choose to move back on premises to gain a greater degree of control," said Seth Robinson, senior director of technology analysis at CompTIA. Over time, this application might remain on premises or the company may move it back to the cloud if they are able to add the appropriate security measures, he said.
For Ullah's team, one of the factors that proved critical in the unclouding process was the rigor and success of the relatively seamless June 2016 cloud migration process.
"We spent a lot of architectural time and effort upfront, so that if we needed to deploy v10 somewhere else -- or if privacy laws changed and we couldn't house the data with U.S. companies any longer -- we would be able to use that same automation to move it to another cloud provider."
The APIs and code the Nightingale team built to automate migration of v10 was tested on premises first, to ensure the system could be deployed to the cloud. The team had also leveraged its internal data center for disaster recovery, making sure the database that contained all of the vital information was always replicated in the cloud and in its data center.
Nightingale used Puppet automation tools for the reverse cloud migration process "to treat the moving back and forth as a failover, almost like a disaster recovery scenario, where we had our live, replicated databases in multiple locations and we would just failover the virtual machines so they went straight to our colocation instead," he said. "It was almost seamless; the only impact was a few seconds of downtime while the networks recalibrated."
Lessons learned
During the unclouding process, automation and configuration management were a huge benefit, Ullah stressed.
"We spent a lot of time building infrastructure as code ... upfront to write software that would deploy an instance of the product in a way that was repeatable," Ullah said. "Once you're able to do it once, it was matter of running the code and changing the variables of where it's being deployed, and it becomes repeatable."
As a result of the v10 experience, today NOD continues to run at TELUS Health in a private cloud but runs using a lot more automation.
"Most of the learning we've gained building v10 we've backported into that, meaning we've added a better metrics collection and automation in tasks [like] deploying upgrades,'' Ullah said. "The automation code is transparent and doesn't care where it goes ... it can be installed wherever."
As for v10 version? It was eventually sold to a U.S. company Ullah declined to name.
Next Steps
Service providers weigh in on reverse cloud migration process
Artificial intelligence advances could hinder Salesforce cloud migration
