Moving large-scale enterprise operations into the cloud is not a decision to be made lightly. There are engineering and financial considerations, and the process of determining the costs pros and cons of such a move is significantly more complex than simply comparing the expense of running a workload on-premises or in a public cloud.
Still, the trend is toward businesses making the move to one degree or another, driven by the easy ability to scale up or down depending on the workload and paying only for the infrastructure resources they use, not having to put up the capital expense to invest in hardware or the need to integrate it all once it’s on site, or the access to the ever-growing numbers of services available in the cloud. At the recent AWS re:Invent conference, the cloud provider highlighted a number of new enterprise customers that have decided to move significant parts of their businesses into the public cloud, including massive online travel site Expedia and high-resolution satellite imagery vendor DigitalGlobe. Expedia is standardizing on AWS to leverage the cloud provider’s machine learning technologies to improve the customer booking experience across not only Expedia.com but also other brands, including Hotels.com and Egencia.
In August, FICO, an analysis and data science company in the financial services vertical, said it was moving key applications – including its Decision Management Suite analytics platform and its MyFICO.com online consumer site. At the helm of the migration was CIO Claus Moldt, who came experience with such cloud companies as Salesforce.com and eBay to FICO when he was hired more than a year ago. The 51-year-old company – best know for its FICO Score credit rating – for a long time has used IBM mainframes in its datacenters, and also uses other systems running Unix and Linux applications. However, the trend for it and for other financial services companies is to the cloud, Moldt told The Next Platform in a recent interview.
FICO currently has a relatively small footprint in AWS – it’s running tens of thousands of virtual machines, a fraction of the number running in its on-premises infrastructure – but there was a lot of planning, validating, test and cost evaluations that happened before a final decision was made, Moldt said. However, the company – which also is still running its private clouds – is already seeing benefits, including being able deploy applications about six times faster in AWS than in its own private cloud, and that could grow to 10 to 15 times in the future.
His time with Salesforce.com and eBay has given Moldt a view of the challenges enterprises face as they migrate some or all of their operations to the public cloud.
“Obviously there’s a difference between companies that are born in the cloud and what they have to go through, versus the companies that were born building software on-prem and the transition they have to go through,” he said. “When you start building for the cloud from the get-go, you start thinking about building at scale and how you can scale up and you learn how to scale up very, very fast. Interesting enough, there was a lot of learning the company had to do itself, and the more recent set of cloud companies have already learned themselves. For both eBay and Salesforce, they experienced such rapid scale and the infrastructure itself sort of came as an afterthought. For both of these companies, you’ll probably find they started out in a very vertical-sale type of infrastructure, where they’re building on very big boxes at the time and quickly realized that we had to go onto commodity hardware and software and do a horizontal software-scaling model across the board. So it’s really about how you can distribute and scale all of your software in a horizontal model that allows a part of these horizontally-scaling tiers. We can afford to use part of that infrastructure and still run, and then obviously building it out so you can replicate data at a fast pace whether that’s done at a cache layer or database layer across multiple disparate regions. You can also avoid full failures on a single region.”
While the trend for many companies is moving to the cloud, how much to migrate to the public cloud and how much to keep behind the firewall continues to vary. The consumer industry has moved more quickly to AWS and other public clouds, but issues enterprises once had about the cloud are disappearing.“A significant amount of workloads are moving to the public cloud,” Moldt said.
“There has been a concern historically about the security and the compliance controls in the public cloud. I think that is fading very, very fast as the public clouds have implemented a lot of the best practices. What people were finding was a set of services that was provided in the public cloud just three or four years ago may not have been on par to what you could find in your private cloud, but that is also fading rapidly. If you just look at the number of services that have been deployed on AWS, for instance, for the last four years, that is just a hockey stick for services. What that truly means for you as people re-evaluate how they think about the public cloud vs. the private cloud is, it’s really about time-to-value for your clients.”
If you run a private cloud, what people will be thinking is, ‘How do I API-enable that private cloud to move faster and deploy faster?’ But you still have the pain of having to do capacity management and doing the procurement and setting the basic things up, which add zero business value. The only thing you can think about from a business-value standpoint is that you would truly have the know-how around security and how you run things. But the fact [is] that you have seen so many resources pushed into the public cloud in terms of understanding both security and order controls. A lot of the services in AWS are already PCI-compliant and already have gotten the stamp of approval, which only makes it easier for people to just focus on their own IP, and then they can point the extension of the PCI compliance in the public cloud.”
A key for moving workloads to the cloud at scale is making you choose the right cloud services for the right workloads. Applications that generate high transactional volume have different requirements than those that create fewer transactions per second. Doing so “helps keep the cost in check. FICO does cost modeling that looks at the costs of running a workload in the public cloud vs. private cloud – how many resources it takes to send a workload out, test and run it, how frequently it needs to be up, the redundancy required, HA [high availability] and DR [disaster recovery], whether local or across regions. Most of services are there; it becomes cheaper to run in public cloud because of scale and capacity.”
That’s not true for every workload. There are some that are very new or may not have been built to scale, and for those applications, an on-premises infrastructure may run them faster, Moldt said. But for increasing numbers of workloads, the cloud offers advantages. Making a financial argument for the cloud takes work. That includes understanding your infrastructure and making a complete review of the cost components.
“That will include everything from manpower to the power and storage space consumption that it takes up, so you have to understand it really in terms of what the breakdown is with all the resources that are required to keep in the internal infrastructure running, whether that’s on-prem or a private cloud,” Moldt said. “Then you have to target the public cloud and you cannot do a lift-and-shift. If you do a lift-and-shift, then you’re going to get your math wrong. Then it’s actually going to look like that it’s going to be cheaper to run in your internal infrastructure, I can almost guarantee you that. If you don’t rewrite some of your applications or rethink how it’s done to take advantage of the feature set that you actually can’t do in a private cloud – things like scaling up and scaling down – that’s a very big impact in the public cloud. Even if you scale down in the private cloud, you still consume the space and the power. You don’t do that in a public cloud, so if you don’t take advantage of the feature set in the public cloud in order to make it economically viable, then you often see that it may not be such a good value proposition to go to the public cloud. That’s an extremely important part how you think about your infrastructure that you can truly understand.”
Private infrastructures come with fixed costs – an enterprise pays for the hardware and software, and often has to overprovision to ensure the demands of workloads are met. In the cloud, infrastructures can scale up or down and customers only pay for what they use.
“Many have started with a lift-and-shift almost: you just set up your instances, you let them run,” he said. “In other words, you treat it as you would run a private cloud or run your private on-prem infrastructure. You have to think differently, and that’s an educational part.”