Oil and gas giant BP is planning on shutting down its two European data centres and shifting 900 applications to the cloud with Amazon Web Services (AWS) over the next two and half years, as part of an ambitious cloud migration strategy.
After already shifting the majority of workloads from its Houston, Texas, data centre to both Microsoft Azure and Amazon Web Services public cloud infrastructure, BP has decided to go all-in with AWS for its European cloud migration.
“We’re a dual-cloud company. For our US data centre, we use Azure as well, and we are big partners of Microsoft. We find having both really helps us, because we have choice. We think both keep each other honest and we keep seeing each leap over the other,” Stewart Fry, global VP of enterprise IT services and digital platforms at BP told Computerworld last week during the AWS Re:Invent conference in Las Vegas.
So why go all-in on AWS for its European migration? “We find when we are really engineering solutions, for the engineers who want to build together, AWS is the preference,” he explained.
Fry feared that by going multicloud in this instance that the company “would dilute down the scale and the help we would get from a partner, and we know we need help from a partner if we didn’t go all-in,” he said.
“We are still dual-cloud for new things, and our teams can choose, we don’t dictate—we don’t say you can go Google—but you can choose either/or. We don’t push one ahead of the other.”
BP turns its back on hybrid
This is all part of a company-wide cloud-first strategy which started in 2013 and advanced from a hybrid cloud strategy to an all-in public cloud strategy in 2016, kicking off a migration plan which will see the company eventually close all of its own data centres, starting with Houston in 2016.
“What actually happened in our first data centre was we didn’t allow any lift and shift,” Fry said, “everything had to go through some level of transformation ... that could be an OS upgrade, we could move the database to [Amazon] Aurora or RDS [relational database service], or it could be a re-engineering.”
Read next: BP’s public cloud strategy helps reduce IT spend by 40 percent
By taking this approach, BP is forcing itself to assess the viability of any applications, creating some welcome natural deprecation of legacy systems. “We rationalised just under 30 percent of our applications out of Houston,” Fry said. “We thought we would do between 15 to 30 percent.”
Fry admits that the challenge is bigger with the European data centre, though, which is BP’s largest and also hosts “our most complex applications in it”.
Houston, we have a problem
The Houston data centre migration will have taken roughly three and a half years to complete when it is completed soon. Now Fry wants to migrate the two bigger, more complex European data centres—which are owned and operated by Global Switch in London—in two and a half years.
“Why do we think we can go faster? Mainly because we’ve learned a ton of stuff and we wanted to be courageous and bold to get ourselves shifted,” Fry said. “And we don’t want to get caught in a refresh cycle of our data centre, because then the business case really does fall apart.”
The key for Fry is to stick to the timeline and reduce any potential impact on the business during the migration. “If we didn’t actually have a business, we could do it a lot faster,” he added. “The problem is all these applications are running something critical or doing something useful.”
For example, BP recently moved one of its bigger SAP environments to AWS, shifting the 16 terabyte behemoth—which controls fuel supply globally, barring the US—in 60 hours. BP already reduced the cost of running its SAP applications by a third after migrating US workloads to AWS in 2017, according to CIO of Downstream, Claire Dickson.
“We’ve moved 2,000 apps now and have enough experiences. We have seen most things, so we know how we would respond to certain parts,” Fry said.
Of course, they don’t have all the answers, and Fry is keen to press home that there are no silver bullets when it comes to cloud migrations. “We have some applications we do not know the answer to; they are on technology today that is proprietary and there isn’t a equivalent version. So the things that keep you awake at night there will be where we will ask AWS and others to come and help us find a solution to that, and we don’t know. But we feel confident that in two and a half years we can solve those problems. Vendors come and see me every week telling me they have a silver bullet and it’s nonsense. We have seen nearly everything and at the end of the day, the big complex stuff comes down to smart engineers who can solve problems with great partners.”
Lessons learned so far
Other key lessons learned from that exercise included moving the big, complex applications first, not last; sticking to the timeline and deprecating applications where possible.
“How do you do the bare minimum to meet your obligations in your data center at the same time as not becoming over invested? That’s why you have to be really committed to the timeline. You can get that quite wrong. That’s a critical piece,” he said.
Naturally, the shift from capital expenditure to operational expenditure for a company like BP requires some cultural change in the finance department, and although Fry didn’t want to go into the nuts and bolts of the finances, he admits that “when you move that to cash you have to try and think through how you explain that”.
Read next: What we learned at AWS re:invent 2019
Fry also admits that it is daunting for enterprise customers to try and leverage all the new products and features coming out of the big cloud vendors. “One of the challenges we’ve got is how on earth do you stay on top of all the new things? Absorb it, make it available for folks to exploit it, and then actually exploit it,” he said. “The speed is just so fast and it is amazing, but as an enterprise consumer it is really difficult to stay on top of that.”
The answer, for BP, has been a hub-and-spoke model, where Fry’s centralised team sets guard rails but departments get to pick and choose the best tool for the job—as long as it isn’t a Google Cloud solution, as he made clear earlier.
“When we first moved to the cloud, we had this really controlled environment, we stretch the VPN out around AWS and Azure and said these are the rules to work within. After a while, we found it was just constraining, because my team couldn’t keep up with the requests for the features and the enablement the application teams wanted,” he explained.
Now, “we set rules in the middle with a few things that really matter: security, versions, some billing; those things will hold the centre. Some application teams still want us to manage their stuff, but the key is to allow people to innovate on the spokes.”
For example, BP is looking to use Amazon’s streaming data tool Kinesis to get nearer to real-time insights for its emissions monitoring and petrol station pump operations on the retail side of the business. It will also look to use the SageMaker machine learning toolkit to develop models to better predict usage and find efficiencies, as well as IoT enabled predictive maintenance of assets.
Sustainability goals
The exclusive nature of this partnership also comes with some shared sustainability goals for the two companies. This includes a commitment from BP to start supplying AWS’s servers with 170 megawatts of renewable power each year, starting with a solar farm in Spain and a Swedish wind farm in 2021, with an expectation to eventually double the capacity in excess of 400MW.
Robert Lawson, chief operating officer for global gas at BP Supply and Trading, said in a statement: “BP and AWS are both targeting reductions in emissions from their respective operations. One way BP can play an important role in helping our customers is by using our trading capability and scale to deliver innovative, reliable and flexible supplies of low-carbon and renewable power to major corporate customers and partners.”
AWS paid for Computerworld UK’s travel to the Re:Invent conference.