As a DevOps engineer, you need to be familiar with various software deployment strategies and know when to use which one. In this article, we’ll look at what software deployment strategies are available, how they work, and the typical strengths & weaknesses of each.
In software development, a deployment strategy is a set of instructions that dictate how our software code or applications should be transferred from one environment to another during the software development life cycle.
What is a Release
The process of "shipping" new features or bug fixes, usually more than one, to users is known as a software release. A software release can be a patched version, a major new version, or a hotfix for an issue found in a previous version. Software releases go through several development stages before they are ready to be made available to users (in what is called production).
A typical software development life cycle includes the following stages:
- System, Integration, and User Acceptance Testing
Your deployment process defines the rules and steps of how software code should be moved (or deployed) from one stage to the next. It is important to have a well-defined deployment strategy because it will help ensure that code changes do not break the software in production and that users always have access to the latest version of the software.
To complete this important job, the DevOps team incorporates deployment procedures into their day-to-day operations. Various approaches have been developed throughout time to help software companies with application deployments.
What Is a Deployment Strategy?
A deployment strategy is a technique used by DevOps teams to launch a new version of their software. These strategies cover how traffic is transitioned from the old version to the new version and can influence downtime and operational cost. Depending on the company's specialty, the right deployment strategy can make all the difference.
Various Types of Deployment Strategies
There are several types of deployment strategies, each with its advantages and disadvantages. The right strategy for your company will depend on your needs and goals.
1. Blue/Green Deployment
This type of deployment process involves maintaining two identical production environments—one is the “live” environment that serves customers, while the other is the “staging” environment. When it’s time to release a new version of the software, the staging environment is switched to live, and vice versa.
- This strategy minimizes downtime because there is always a production environment available.
- However, it can be costly to maintain two identical production environments.
2. Canary Release
In this strategy, the new version of the software is first released to a small subset of users. If there are no major issues, the new version is then gradually rolled out to a larger subset of users until it is finally made available to the entire user base.
For example, the older version may retain 90% of all traffic for the software at a certain point in time during the deployment process, while the newer version hosts 10% of all traffic. This method helps DevOps engineers to test the new version's stability. It utilizes real traffic from a fraction of end-users at different phases throughout production.
- Better performance monitoring is possible with Canary deployment. It also aids in the quicker and more successful rollback of software if a new version fails.
- However, it does require more effort and typically, a long deployment cycle.
3. A/B Testing
May, also be called Incremental Rollout
In the A/B testing deployment process, developers deploy the new version alongside the older version. This type of testing is used to compare two versions of a software feature to see which performs better. Version A is the control and is made available to the entire user base, while version B is the test and is only made available to a subset of users.
A/B testing has several deployment process benefits:
- It allows software developers to compare two versions of a software feature to see which performs better.
- It is easier and less risky to test a new version of the software on a small subset of users before rolling it out to the entire user base.
- Developers can easily accept/reject either version.
- Increased user/customer coordination.
4. Feature Toggles (Feature Flags)
Feature flags are a type of deployment strategy that allows developers to turn on or off certain features of the software for different users. This allows developers to test new features without making them available to the entire user base. Feature flags can be used in conjunction with other deployment strategies, such as A/B testing, to help developers test new features before
5. Recreate Deployment
In this deployment approach, the development team completely shuts down the old software, then deploys and reboots the new version. This method causes a system outage between shutting down the old program and booting up the new one.
- It is less expensive and is primarily utilized when the software company wishes to rewrite the application from the ground up. There's no need for a load balancer since there are no changes in traffic flow in the live production environment.
- This method has a significant impact on end-users since it is unavailable/suspended. Users must wait until the software is reactivated before using it. As a result, few developers employ this technique unless they have no other option.
6. Trunk-Based Deployment
In this strategy, all code changes are first merged into a main trunk or branch. Developers then create a new branch for each new feature. Once the feature is complete, it is merged back into the main trunk. This strategy eliminates the need for long-running feature branches and makes it easier to deploy new changes.
Note: This is more a pre-deployment method of Software Version Control.
7. Ramped Deployment
The ramped deployment method moves from one version to the next in a gradual process. Unlike canary deployment, which replaces instances of the old application version with those from the new application version one at a time, the ramped deployment approach makes its change by replacing instances of the old application version with new applications. The rolling upgrade deployment strategy is another name for this method.
The second method, as the name implies, is to delete the old version from production. When all of its instances are deleted, the older edition is manually shut down. The new edition then controls all production traffic.
- No need to take the entire application offline for an upgrade.
- The process is gradual, so it's less risky.
- Takes longer to complete than other methods.
- Requires more instances to be available during the process.
- Rollback is more complicated & long.
8. Rolling deployment
For those using containers.
Rolling deployment is a gradual process of replacing pods running the old version of the application with the new version, one by one, without downtime to the cluster. It is less risky and takes longer to complete than other types of deployment, but it doesn't require taking the entire application offline.
- Lower Risk
- High Availability
- Only really applicable for container-based architectures.
9. Shadow Deployment
Developers deploy the new version alongside the existing one in this deployment method. Users, on the other hand, won't be able to access it right away. The newest version hides in the shadows, just as its name implies. Developers send a fork or copy of the previous version's requests to the shadow version so they can examine how the new variant will work and if it can process the same amount of requests.
When the shadow version can handle the same load as the original, the traffic is finally routed to the new version, and it becomes live. The cutover from the original to the new version happens without any significant downtime since there's no need to take down or restart either version.
- valuable feedback can be gathered about how the new version will work in production
- there's no need to take down or restart either version during the cutover process
- more complicated to set up and maintain than other deployment strategies
- if not done correctly, it can cause issues with the live version
When to use:
- when you want to gather feedback about how the new version will work in production
- when you want to avoid any significant downtime during the cutover process
Deploy Better with a Software Deployment Tool
Managing your deployments without tools can be fraught with danger.
As seen above, the different deployment processes can be quite fragile/awkward, and if done incorrectly could lead to production issues, outages, and the need to roll back.
Using tools to control your "implementation day events" can uplift visibility, improve collaboration, support rehearsal, standardize your operations and also streamline the tasks*.
*Tasks that may be manual or preferably automated.
Fortunately, there are various Release Management tools that can help your organization with the various aspects of Environments, Release Management & Application Deployment.
The best software deployment tools included features like:
- Release Management Governance for Scale Delivery*
*for managing the End to End Release / Release Train.
- Implementation Plans (for Deployment Planning)
- Operational Runsheets / Standardized Operating Procedures
- DevOps Automation e.g. Software Deployments
- Orchestrations / Integration with other tools*
*deployment tools, ticketing tools, CI/CD i.e. continuous integration, and continuous delivery tools
- Deployment Version Tracking
*tracking code deployments across Environment Instances, Components & Microservices.
- Environment Drift Reports
*supporting holistic, cross-environment, version control
You may use any of these methods to upgrade your applications. Each of these approaches has advantages and disadvantages, and each is appropriate in certain circumstances. The only question now is which one makes the most sense for your DevOps team to utilize.
Consider the demands of your team, project, and company as well as corporate objectives. Also, keep track of how much downtime your business can tolerate and any other cost limitations.
Make your go-live events into non-events!
Uplift your Implementation Planning, and Deployment Management capability today. Find the best software deployment tool (or tools) to help with your automatic deployments.
Author: Mark Dwight James
This post was written by Mark Dwight James. Mark is a Data Scientist specializing in Software Engineering. His passions are sharing ideas around software development and how companies can value stream through data best practices.