Mastering Cloud Migration: Best Practices to Make it a Success

Cloud migration projects are picking up pace as top executives are pushing for more efficiency and cost savings. And it may be tempting to say that cloud is the answer and that it’s the way to go for everyone. It may be tempting, but it’s not quite true. Throughout my professional career as an engineer, I’ve seen different scenarios, from middle-size companies deciding to move to the cloud to enterprise-grade corporations making a choice in favor of on-prem infrastructure. And I know that path is anything but straightforward.

If your team is getting ready for cloud migration, this article will provide you with an actionable approach to the task at hand.

To Cloud or Not to Cloud, or 3 Questions To Answer Before You Build Your Cloud Migration Strategy

The story of cloud vs. on-prem is often seen through the lens of convenience vs. control. With the cloud, you can choose the operating system and manage patches, but you can’t really control the hypervisor or hardware performance. So, whatever path a company decides to go, there must be clear-cut reasons and objectives that the business is trying to achieve.

I suggest answering the following three questions to ensure that cloud migration is the right move.

1. How Critical Is Real-Time Performance for Your Application?

Solution performance is always important; there is no arguing with that. But for some network-intensive applications, real-time performance becomes mission-critical. Think of an ad-tech platform where latency is the biggest enemy – anything above 100 milliseconds can translate into lost bids, ad space, and revenue. Thus, any delay caused by the network interface, core, or driver can become a serious problem. Your on-prem data center is the only place where you can truly control those things by changing the system core, installing necessary patches, and changing the driver or the network interface to get maximum performance.

2. Do You Have a Heavy but Highly Predictable Workload?

Now, let’s consider compute-intensive apps that handle large-scale data analysis, statistics calculation, machine learning model training, and other tasks that require high computational power. With these types of solutions, your hardware is tirelessly working almost at maximum capacity. And the cloud computing cost of running these CPU-intensive apps can be extreme.

3. Do You Have Seasonal Spikes in Demand?

The problem with on-prem data centers is that you can’t immediately scale up if you experience a spike in traffic. To do that, you will first need to physically add new hardware. But in the cloud, scaling up or down can be done within seconds with no disruption to business operations. So, if you regularly experience seasonal demand spikes (e.g., before Christmas), the cloud will give you the necessary flexibility and agility to meet that demand.

How To Approach Cloud Migration

Now, we have come to the most interesting part – moving your system to the cloud. If you have a Salesforce-based application, it is easy to find a cloud migration tutorial on the Internet. However, if your ecosystem consists of a Salesforce platform and a custom-built solution, all connected through tailored ETL processes, there can be no ready-to-use guide. So, keeping in mind that there is no one-size-fits-all approach, what do you start with?

1. Perform a Baseline Assessment

Quite logically, the first step would be to understand what you are dealing with – what technologies and databases your solution uses and what other systems and components it is integrated with.

We have already talked about how the cloud takes over the control of your solution’s operational environment. So what if your system directly interacts with the cryptography board or uses the license linked to the hardware ID? It is essential to understand early on what elements and integrations are critical for the overall system operation. 

2. Prioritize Migration Queues

As we have discussed earlier, cloud migration, just for the sake of it, is never a good idea. There must be valid reasons and tangible goals a company wants to achieve with this project. At the same time, during the preliminary baseline assessment, we have uncovered where the greatest problems during migration can arise. When we match these things, we can prepare an effective migration plan that would prioritize the most important issues. For instance, if the main objective of moving to the cloud is to resolve scalability bottlenecks, then the system that needs scaling will be the first in line to migrate to the cloud.

3. Map Out Dependencies

Once you know what to move to the cloud first, you need to understand what to do with dependencies. Let’s say your application uses three databases: one database is used exclusively by your app, so it’s easy to migrate it to the cloud, but the other two are shared with other systems.  The question is, what are you going to do with these databases, as well as the systems that use them?

Well, there are different options here. In some cases, you can set up a shared network infrastructure by extending your on-prem data center to the cloud through a VPN. Or you can migrate a database to AWS, and your solution will easily retrieve data from there. However, data migration to the cloud can be difficult if you have lots of it. Also, since you can’t disrupt the system’s operation and related business workflows, you will need to spin up and configure a replication instance, and only then can you set up migration queues.

4. Choose the Right Strategy

The simplest cloud migration strategy is to lift and shift. If your solution is hosted on a virtual machine on a server, the easiest way is to take a VM image and just transfer it to the cloud. However, there is one caveat to it – this strategy incurs significant operational costs. When you launch a virtual machine, you pay for the entire time of its operation, no matter whether it actually runs or not.

If lift and shift is not your option, you can break the system down into components. For instance, if your database is running on a virtual machine, you can move it to a new virtual machine in the cloud or use managed services for it and then reconfigure the connection string for this database. As a result, you not only migrate your database to the cloud but also boost availability as Amazon services run in different geographical areas.

Another way to approach your cloud migration strategy is to look at whether your solution is a monolith or has a client-server architecture with a front-end client app and a number of APIs. If yes, then these APIs can be hosted on separate virtual machines. Or you can also refactor the backend to Lambda functions, moving a step further to the serverless architecture.

Wrapping Up

As it is clear from today’s discussion, cloud migration is a challenging yet fascinating topic. And since every solution is unique, with its own tech stack and dependencies, there is no one single guideline or tutorial for cloud migration. However, there are best practices – do not neglect your baseline assessment, prioritize migration queues in line with business needs, identify all dependencies, and prepare the roadmap best suited to your project.