Cloud migration: the main challenges and solutions

What is cloud migration?

Cloud migration (CM) term speaks for itself: it refers to the processes of moving, migrating, relocating of the whole app (or a project—in the broader meaning) from on-site, local, bare metal servers and storage to cloud-based providers. They can be public (AWS, GCP, etc.), private (like IBM Cloud Private, VPC), or combine deployment models, both public and private, in a hybrid cloud.

There are different types of cloud migration. CM can be used for description of cloud infrastructure migration to another cloud. Companies might change cloud providers occasionally to save resources and increase stability.

By 2022, over 90% of companies worldwide will be relying on a mix of on-site/dedicated private clouds, multiple public clouds, and legacy platforms to meet their infrastructure needs. IDC Report

Another trend of a multicloud — using few cloud providers (like AWS + GCP) and storage services at once—based on hybrid clusters is popular because of increased fault-tolerance and availability, reduced servers' latency, and a better user experience. In short, that's all about the various cloud migration types.

Why businesses become interested in cloud migration solutions

• Access to advanced technical appliances: public cloud providers have more expensive and innovative equipment unavailable even for large companies.
• Improvement of stability and fault-tolerance of the app by means of a high stability of cloud migration service providers. Bare metal servers of separate companies are less stable than servers of cloud providers because of unevenness of maintenance efforts and costs. To reach the same level of stability, you need to invest in complex infrastructure and its constant improvement.
• Quick scaling with minimal efforts. A cloud provider will allocate you as many resources as you need, and cloud migration costs will be calculated automatically. You'll pay for them only when you use them. The efficiency of resource usage is a reason for drastically reduced costs, a dream of every business.
• Out-of-the-box cloud solutions for business and only-in-cloud available tools. SaaS cloud services have a great variety of ready-made and full-stack solutions for your projects for quick adaptation and performance in the cloud environment. Do you need a cloud-native database solution? AWS RDS or GCP Cloud SQL—at your service. Or do you dream about the possibility of using GCP AutoML or Azure GPT-3, but those aren't affordable even for large companies? In the cloud, you will find an enormous toolbox with instruments created specifically for tasks of automation, connection, security, and stability. That's why you should always consider any cloud migration options for your project.
• Sustainability goals and upcycling of data centers for limiting the environmental impact. It's true that company data centers are not the most safe and eco-friendly places. A social responsibility principle brings new reasons for companies to move to the cloud, thus minimizing the carbon footprint and becoming more sustainable in their development, just like the Capital One case.

Challenges of migration to cloud services

Generic problems

Here are some of the most common intricacies businesses encounter before or even during the cloud migration process. We prepared them in the following list of advice and "things to consider":

1. Common misconceptions of the role of a cloud within the heads of management and developers team:
2. "We're not going to do everything like before and the cloud."
3. "The cloud can be just another copy of current infrastructure."
4. "Our current infrastructure is working well enough, so why would we risk using it for any type of cloud migration?"
5. All of these thoughts are demonstrating a lack of knowledge about CM, and maybe you should consult a specialist to clear these things up.
6. Corporate culture shift, reinventing of the corporate digital structure and a change in mindset:
7. "People go first, tools — later": invest in the right specialists and ask them what tools are better for your company rather than investing in tools, not vice versa.
8. Make all technical and non-technical departments understand what's going on with the company and how it affects their processes.
9. Anything new will be met suspiciously if you don't explain how and why you're going to implement it. Find the people from a migration-as-a-service sphere who'll become advocates for a cloud migration and explain why everyone wins from this change.
10. Cloud data migration services: transferring of persistent storage to object storage:
11. When any of your data is stored in files, your scaling depends on hardware storage volumes. There would be no use for this analogical structure in the cloud.
12. The correct decision is the migration of all static data into Object Storage (Google Cloud Storage for GCP, S3 for AWS).
13. The benefit of this solution is that you won't even think about scaling your memory storage again, even if your data increases drastically. You'll be able to store various types of data for various purposes safely and securely, and pay less for data that you don't require quick access to.
14. Legal and data privacy issues can cause concerns about cloud migration security:
15. The storage of sensitive data for some business domains such as health care or fintech requires a certain level of protection of the on-prem infrastructure, sometimes even based on some legislative reasons.
16. If there are no unavailable prohibitions, moving to the cloud is a way to reconsider your security and improve it.

The big three cloud providers (AWS, GCP, Azure) are following the “secured-by-design” concept and include a range of services such as:

• Encryption of the stored and transferred data;
• KMS, AWS Secrets Manager for personal data management;
• Centralized systems of authentication and access rights managements (IAM);
• Tools for cloud services access (like SaaS databases) from certain compute instances without a need to store credentials (GCP Workload Identity);
• Organization of structures that comply with most popular security standards (e.g., PCI DSS).

All these concerns can affect or even prevent the start of cloud migration projects, but the biggest fish in the lake is still up ahead. Let's proceed to the biggest challenge of CM and how Dysnix typically overcomes it within the optimal average cost of cloud migration.

Major challenge: real-time cloud migration without downtime

Don't expect random work from random vendors to magically work together.Pieter Hintjens, Confessions of a Necromancer

Unplanned and unsupported migration can take too long and affect the business process, as the case studies of the best cloud migration companies tell us. Time is always the most valuable resource, and downtime because of migration will grow your bounce rate, and it'll affect your income. To avoid it, use the key for successful cloud migration: plan every step of migration for each data layer you've already had:

Cloud database migration

The most efficient method of cloud data migration without loss of information and downtime is to set up the replication from the old environment to the new. If you have a high load on your database, it's better to use synchronized replication for keeping your cloud database identical to your main one. It'll slow down the database response for clients a bit, but you'll be sure that none of the transactions were lost in the old environment after the start of migration.

Object storage

Aside from self-hosted solutions, it's typical to use Google Cloud Storage or AWS S3 for Object Storage tasks. In this case, nonstop migration require some changes in the code of your app:

• Add a new cloud location for files being saved in your app code, so now you have both new and old locations saved.
• All "old" files exiting in bucket before double upload setup can be copied with standard utilities like AWS CLI.

As a result, we'll have two identical and updated storage. After all files are checked in the cloud storage, the function of saving to the old storage is turned off.

Switching app traffic without downtime

After migration of the existing data and updating it is over—and the app is replicated in the new environment and tested there—it's high time to test everything with real traffic. But how can we do it with DNS TTL that doesn't allow us to switch all the customers simultaneously? Or what if you get into trouble over things that weren't covered by tests and need to switch back quickly without any discomfort for your client?

At Dysnix, we use the "divide and conquer" principle for such risky procedures and create step-by-step action lists to minimize any complications in the implementation of cloud migration technologies:

1. Switching traffic to the outer Cloud Load Balancer (CLB). This tool will be a “single window” for managing all your traffic in the future. In the beginning, your CLB will point to your old servers, and for your customers nothing will change. Later, we test this load balancer precisely and switch the DNS while waiting a reasonable amount of time until all traffic leaves the old endpoint (2-3 days, typically).
2. When we implement the control of the traffic routing, our engineers use A-B methodology of deployment and switch 1% of traffic to a new environment on CLB level. At the same time, we revert to the old endpoint as a backup. In the other words, we allow only a small amount of traffic and analyze how the environment works, but if something goes wrong, all customer requests will be connected with the old environment seamlessly.
3. When we see that the new environment is working reliably with 1% traffic, we gradually increase the traffic load up to 100% and check the logs and monitoring indicators. The old endpoint still works as a backup.
4. After 2-7 days of 100% load when we assure that everything is working good, we make the final reserve copy, shut the old environment down, and delete the data.