Facilitating the Spread of Knowledge and Innovation in Professional Software Development

Write for InfoQ


Choose your language

InfoQ Homepage Articles Lessons Learned About Cloud Migration

Lessons Learned About Cloud Migration


At the recent Dynatrace Perform conference I did a panel session with a variety of technology companies on the topic of cloud migration. The experts ranged from SaaS companies to eCommerce companies and cloud providers. Company sizes ranged from a startup, to one of the biggest eCommerce providers in Brazil, to an established SaaS business. Although the companies were rather different they had a strong consensus on key facts of cloud migration projects. This article covers the top five findings.

There’s always a compelling event

The panel agreed that cloud migration is not a decision that is made out of the void. For all participants there was a compelling event that made them make the move to the cloud. For B2W the compelling event was the rearchitecture of their entire environment to microservices following a serious production incident. Moving to the cloud was a key part of the strategy as dynamically scaling individual components based on traffic was a key part of their cloud strategy. For BARBRI– a leading provider of only BAR exams - the key driver was a datacenter renewal that made them decide to move to the cloud.

Costs are a key constraint when moving to the cloud

The first thing that comes up when asked about the motivation to migrate to the cloud was cost reduction . All participants agreed that reducing operational costs was the key driver for moving to the cloud. For the smaller startups, the elastic model of IaaS enables just-in-time investments into infrastructure as needed. For established companies, a key event is having to reinvest in datacenter renewal. Once you face a big investment to make, you rethink your strategy. “For me it was the top line cost that we spend on our datacenter today” said Mark Kaplan, Director of IT at BARBRI. An important point to note is that cost benefits take a while to emerge. In the transition phase companies run their cloud and non-cloud infrastructures concurrently. “You still have to plan for cost savings with a two-year horizon to make a project successful”Kaplan pointed out.

Placing APM tools on the CIO agenda

An interesting side effect of cloud migration is that it placed APM tools on the agenda of the CIO. With the elasticity and flexibility of the cloud, companies can more directly and immediately drive cost optimizations. Greg Birdwell from BARBRI made the point, “We use an APM tool not only to monitor the health of our infrastructure. If I see that we have servers consuming significantly less CPU or memory, I can switch to cheaper instances. The cost savings are immediate.”  Mark Kaplan of BARBRI said that Ruxit helped them to get exactly the insight into the dependencies and resource requirements of their environemt as basis for migration  It might still take a while until we see CIOs around the world looking at APM tools and calculating cost benefits based on monitoring data, but things are moving that direction.

Agility beats cost benefits

One of the key surprises of the panel was that although cost reduction was a key driver for cloud migration, all agreed that they are now spending more money on cloud infrastructure. “It would be cheaper for us to run a classic datacenter” said Michael Aigner of Stilnest. “We are still ‘all in’ on the cloud though because we get something more valuable than saving a few bucks: agility”. All panelists agreed that cloud services help achieve faster time-to-market by supporting agile development of new functionality, which is appreciated by customers.

Cloud migration impacts your culture

The participants agreed that cloud migration is not a technical or economic game changer, but it broadly affects company culture. If you haven’t made the move to DevOps already, you will once you run in the cloud. “All our infrastructure is now in code” said Michael Aigner of Stilnest. “I expect our operations team to become more like and work like developers”, said Alexander Ramos from B2W. Developers are continuously involved in resolving production issues and, according to Anita Engleder of Ruxit, the DevOps lead is often the enabler for development teams becoming part of application operations. “They know more about their code than I do anyways, so they should be responsible for monitoring” Engleder concluded. Alexander of B2W took this to an extreme: “I have handed out pagers to the developers, they are on-call like everybody else. This has changed a lot.”

Selecting a cloud provider can be a challenge

The panelists agreed that selecting the right cloud provider is a key challenge. Picking a cloud provider depends on your infrastructure needs and the flexibility of the cloud provider. However, there is more than just infrastructure to consider when evaluating price points. “If you need something, you want to know that somebody is answering the phone. Reliability and a trusted partner is key for every cloud move,” said Bob Stolzberg of CenturyLink Cloud.

Successful migration is based on rigorous planning

Bob Stolzberg of CenturyLink pointed out that executive sponsorship and end-to-end planning are key for successful delivery. Based on a large set of migration projects, his advice is pretty clear: “Have a plan for execution and shoot for success, but not perfection” Stolzberg said. “You also need to have a rollback strategy. If things go wrong, you want to be on the safe side.”

Cloud migration doesn’t mean moving to a public cloud

While for many organizations cloud migration means moving to the public cloud, this is in fact only one option. Especially when regulatory requirements are important, public cloud are simply out of scope. Romain Bigeard from Avocado Consulting points out that for companies in this case the direction to go is toward a private cloud. “A lot of the benefits expected from cloud environments can be leveraged by private cloud infrastructures as well,” he concluded. For them the key was to build a programmable infrastructure that enables higher agility in their software delivery processes.

Making the right move

Amongst the session’s participants, an informal survey revealed insights into the audience’s cloud-migration strategy. The majority decided to get their feet wet by first moving a new application to the cloud. Once they gained some experience, they began to move existing on-prem applications to the cloud. They did this by either making the move all at once or by first moving only certain application services to the cloud.

Infrastructure-as-a-Service is the most common reason

While cloud providers provide options like Containers-as-a-service and PaaS offerings, the audience still favors Infrastructure-as-a-Service offerings. Still, many leverage additional cloud services to optimize application delivery.

Understanding dependencies is the biggest challenge, followed by sizing

There was wide agreement that the biggest challenge in cloud migration is understanding the dependencies within their existing infrastructure. “Sometimes I wonder why two applications are even talking to each other. There is no documentation, so I have to get all this data out of Ruxit,” said Alexander of B2W. Sizing comes next on the list of challenges. Prizing for cloud instances may vary significantly depending on size, so running the right instance types is key from a cost saving perspective. Traditionally companies bought large enough machines to be on the safe side. In the cloud companies want to buy the smallest possible infrastructure and scale up when needed. Especially in a lift-and-shift scenario the sizing of the current datacenter server is often overprovisioned. As mentioned, infrastructure costs are more immediate and cost savings are key.

Cloud monitoring is different, too

Once you’ve migrated to the cloud, things are different. Having all these new cloud services as part of your application means you need tools that can natively monitor and manage them. Scaling cloud environments comes next on the list of challenges. All the theory behind cost savings by only running what you need may be fine on paper, but automatically scaling your environment hard as tools offer very limited functionality beyond scaling based on resource consumption. Cost savings can be significant ultimately, but initially you need to focus on ensuring that your migration efforts actually work without affecting performance and user experience. Once everything works fine, however, capacity planning in both directions – scaling up and down – becomes a constant exercise. In a world of consumption based pricing models, hours can make a financial difference. 

About the Author

Alois Reitbauer is Chief Evangelist at Ruxit. He has spent most of his career building monitoring tools and fine tuning application performance. A regular conference speaker, blogger, author, and sushi maniac, Alois currently shares his professional time between Linz, Boston, and San Francisco.

Rate this Article


Hello stranger!

You need to Register an InfoQ account or or login to post comments. But there's so much more behind being registered.

Get the most out of the InfoQ experience.

Allowed html: a,b,br,blockquote,i,li,pre,u,ul,p

Community comments

Allowed html: a,b,br,blockquote,i,li,pre,u,ul,p

Allowed html: a,b,br,blockquote,i,li,pre,u,ul,p