Migrating to the Cloud: Top Considerations when Replatforming Applications

Moving Database Applications to the Cloud: How Hard Can It Be? (Part 1)

09.26.18

Why Are So Many Enterprises Moving to the Cloud?

09.12.18

Replatforming is the process of rewriting existing enterprise applications to be compatible with a modern data warehouse, and is the most common approach to migrating applications to the cloud.

Today, many enterprises are looking at cloud databases and warehouses (Gartner believes that the majority of new licensing buying likely to be for cloud deployments by 2020) because of the advantages they offer: ongoing technical innovation, security, lower CAPEX and OPEX, better scalability, less maintenance, and little-to-no tuning. This means many of them will be replatforming soon.

In this post, I’ll talk about the importance of planning and considerations that must be kept front and center before migrating to the cloud. Some of these considerations include assessing current approaches and limitations, the importance of understanding the complexity of existing database workloads before making a replatforming plan, why replatforming is so difficult and expensive, and the benefits of virtualizing the data warehouse.

Migrating to the Cloud: Possible Approaches

There are many ways to move from an on-premise data warehouse to a cloud-native data warehouse, including rehosting, refactoring, replatforming, and repurchasing, among others. Each approach has its own pros and cons, some are easy but don’t support every kind of application, some strategies are more complete but are incredibly time-consuming and difficult to pull off.

  • Repurchasing the cloud-equivalent application: If an enterprise is moving a Microsoft application from an on-premise Azure data warehouse to Microsoft Azure SQL DW, it is possible that the application is already available and supported. In that situation, the enterprise could simply end the current subscription and repurchase the cloud equivalent of the application. The con to this approach is that most enterprises use a wide variety of applications from many different software vendors, and many that they have developed or customized over time.
  • Rehosting the application: This is also referred to as lift-and-shift. It involves moving existing physical and virtual servers as is into a compatible IaaS solution. This scenario works best with applications that are already similar to or supported by the target cloud. Again, this approach will not be useful in every situation.
  • Rebuilding the applications: This would mean starting from scratch, building the application again to work on the new cloud system. This method varies in difficulty, depending on the complexity of the application, and only works for applications the enterprise has developed in-house.

The approaches described above could fit the needs of smaller databases or an application or two but simple data warehouses are not the norm for Global 2000 enterprises: data warehouses are the crown jewels of the enterprise and can contain millions of lines of code painstaking built over decades to serve business needs. For example, one Datometry customer had a 250TB database with a daily average query volume of 14 million. In this situation, none of the three options listed above will work. The enterprise would need to rewrite all the applications line by line for the destination cloud data warehouse and test them: migrating to the cloud would be a multi-year, expensive initiative with costs running into the tens of millions of dollars.

In fact, the complexity and challenge of enterprise replatforming initiatives is confirmed by Gartner which found that 50% percent of all migrations fail, go over budget, or run late because the effort of rewriting and testing applications to work with the new database is severely underestimated.

Data warehouses are the crown jewels of the enterprise, and can contain millions of lines of code painstaking built over the course of decades. Click To Tweet

Understanding the Complexity of Existing Data Warehouse Workloads

Before migrating to the cloud, enterprises need to extensively plan and prepare for a successful application rewrite. A thorough understanding of the enterprise’s applications and creating a detailed plan before replatforming are paramount to success. Mistakes made when rewriting applications could cause downtime and create significant risk for the business. A deep dive into existing workloads includes getting familiar with the application stack and figuring out which applications are already compatible with the new cloud destination and which need to be rewritten.

Typically, a system integrator is tasked to review the applications, assess the business user needs, create the architecture for the new cloud-native data warehouse, figure out how long the project will take, and determine the overall cost. This is a mostly manual effort and can take months to complete depending on the complexity and size of the data warehouse. Another approach can be using a tool or software solution, such as Datometry qInsight to determine the workload compatibility with the target cloud data warehouses. With qInsight, you can receive a detailed report summarizing the functional, operational, and performance characteristics of data warehouse workloads, assessing the compatibility of the workloads with the target data warehouse, prioritization of workload hygiene, and more.

Analyst firm Gartner found that 50% percent of all migrations fail, go over budget, or run late because the effort of rewriting and testing applications to work with the new database is severely underestimated.

Why Replatforming is so Difficult and Expensive

Interestingly, a recent survey conducted by Datometry found that the technical effort and cost of rewriting applications were the biggest challenges in replatforming for the majority of respondents. Where does all this time and risk come from?

Replatforming can be broken down into a few parts. First, enterprises need to develop a good understanding of their application stack and determine which applications will need to be rewritten. Then, the schema will need to be generated for the cloud database or data warehouse. After that, the enterprise rewrites their applications, and finally, the enterprise can move their data into the cloud data warehouse. Most of the difficulty, risk, and cost are centered in the first three steps. It’s surprisingly hard to get a complete view of the workloads and queries that run every day, and of course, the process of generating the new schema often takes at least a few months and the costs can run into several hundred thousand dollars. But, rewriting the applications remains the biggest challenge. In fact, one customer said their plan to rewrite the applications would take five years and cost USD $35M.

Virtualizing the Data Warehouse

New technologies—like Datometry Adaptive Data Virtualization—are making the adoption of cloud databases easy by making the application rewriting process obsolete. Virtualizing the database in weeks, not years, with minimal risk to the business, and at a fraction of the traditional replatforming costs and the competitive advantage offered by modern cloud databases is truly a winning combination!

7 things every IT leader should know before moving analytics to the cloud
Download PDF

About Travis Singleton

Digital Marketing Manager

Travis is a Silicon Valley native with a passion for all things high-tech. From a young age, Travis felt a deep connection to technology, taking part in Lego robotics classes and developing his first website in fifth grade. Presently, you can find Travis at the forefront of technological disruption, working with startups like Datometry to build a path to the future.

Continue Reading