SRE vs AWS DevOps: A Personal Experience Comparison

With hands-on experience in AWS DevOps and Google SRE, I’d like to offer my insights on the comparison of these two systems. Both have proven to be effective in delivering scalable and reliable services for cloud providers. However, improper management can result in non-functional teams and organizations. In this article, I’ll give a brief overview of AWS DevOps and Google SRE, examine when they work best, delve into potential pitfalls to avoid, and provide tips for maximizing the benefits of each.

DevOps

DevOps is a widely used term with multiple interpretations. In this article, I’ll focus on AWS DevOps, which, according to the AWS blog, merges development and operations teams into a single unit. Under this model, engineers work across the entire application lifecycle, from development to deployment to operations. They possess a wide range of skills rather than being limited to a specific function.

As a result, the same engineers who write the code are responsible for running the service, monitoring it, and responding to incidents. In practice, every team may have its own approach, but there is some degree of unification of practices, such as with CI/CD, incident prevention, and blameless post-mortems. Personally, I consider AWS to have the most effective operational culture among all the organizations I’ve worked with.

Advantages of the DevOps Approach

When DevOps is implemented effectively, it can provide several benefits, especially in the early stages of development. For start-ups looking to bring a new product to market quickly, DevOps can offer speed and agility. Similarly, established companies launching a new service or product can also benefit from the DevOps model. 

Although the same team operates the system, there may be some specialization, with some team members focusing more on operations and others on development. Over time, as the product matures, teams may split, with a platform team (akin to SRE) working alongside a development team (akin to SWE). 

However, the integration and overlap of operational activities by the development engineers and deep understanding of the system by the operational engineers remain tight. This tight feedback loop leads to a better understanding of how the system runs, its limitations, and the customer experience by all team members. 

This, in turn, makes decision-making and iteration cycles faster. This is likely a contributing factor to AWS’ dominance in the market and the large number of offerings it provides.

When DevOps Goes Wrong

Generally, operations can be divided into three main categories: 

1. Service operations
2. Incident prevention
3. Incident response

While service operations are often seen as enjoyable by software engineers, incident prevention may not be as engaging, and incident response can become overwhelming, particularly when engineers are responsible for development and operations. The more time they spend on operational tasks, the less time they have for development and the more dissatisfied they become with their job. 

This can result in a vicious cycle of overworked engineers, high turnover, decreased work quality, and a growing workload for operations.

Site Reliability Engineering (SRE)

Site Reliability Engineering (SRE) is a discipline developed by Google to improve the reliability and availability of software systems. It involves a dedicated team of SREs who focus solely on these goals, while software engineers (SWEs) handle writing the code. SRE brings a formalized set of principles and terminology, such as Service Level Indicators (SLIs), Service Level Objectives (SLOs), error budgets, toil, and others, to ensure the software is scalable and meets performance standards.

Benefits of Site Reliability Engineering

When SRE is implemented effectively, it provides a high level of standardization and consistency in measuring customer experience. This approach doesn’t necessarily result in more reliable or performant services, but it ensures that best practices are followed across multiple products. By having dedicated SRE teams, it reduces the burden of operations on the software engineers, who no longer need to deal with operational issues at all hours of the day and night. As a result, software engineers can have a better work-life balance, while the SRE team ensures that operational needs are met in a consistent and efficient manner. 

When SRE Goes Wrong

In the SRE model, software engineers (SWEs) are freed from the operational burden; however, this can result in a lack of exposure to the workings of the system, leading to vague risk assessments and limited understanding of how their code behaves in different conditions. On the other hand, SREs may be overburdened with an excessive number of pages, which can slow down development by becoming overly risk-averse. This, in turn, affects the SWEs who then become risk-averse and struggle to get approvals from SREs. 

This disconnect between the two teams, with SWEs perceiving the service as a black box and SREs lacking an understanding of the code and intent, can lead to a semi-functioning organization where deploying code to production may take months and the majority of initiatives never see the light of day.

Which One Is Better?

The answer is not that simple. Neither DevOps nor SRE is inherently better or worse, they both have their own strengths and weaknesses. 

When it comes to DevOps, it’s crucial to ensure that engineers are not overburdened with operational tasks, and that they have a healthy work-life balance. This can be achieved by proper investment in tooling and a focus on quality output. Additionally, it’s important to strike a balance between development and operations to avoid a situation where either one of the two becomes more dominant and hinders the progress of the other.

On the other hand, SRE is designed to alleviate the operational burden from software engineers and protect them from the distractions of incident management and other operational tasks. However, it’s important to avoid a disconnect between the SWEs and SREs and ensure that each team has a comprehensive understanding of the system. Additionally, SREs should not only be focused on operational metrics, but also be interested in delivery and should have skin in the game.

In other words, both DevOps and SRE have their own advantages and disadvantages, and the best approach will depend on the needs and culture of your organization. The key is to avoid the pitfalls of each system and strive for a balanced and effective approach to software delivery.

Balancing Speed and Stability

Balancing speed and stability is a critical aspect in the DevOps vs SRE debate. The approach that a company takes will depend on its stage and goals. Start-ups often prioritize speed and agility to bring their product to market quickly, making DevOps the ideal choice. As the company grows, stability and reliability become more important to maintain customer trust, making SRE a better fit. 

However, the transition from DevOps to SRE does not mean giving up on the principles of speed and agility. An effective SRE model can still strike a balance between reliability and speed by ensuring close collaboration between SWEs and SREs. The SWEs drive the development process, while the SREs ensure the system is reliable and scalable. Regular hat-swapping rotations and joint operational meetings can keep both teams tight-knit and aligned with delivery and stability goals. This approach offers the best of both worlds solution.

Closing Thoughts 

The choice between DevOps and SRE is not straightforward. The best approach depends on the situation of your company and what it needs. By combining the advantages of both, you can find the sweet spot between speed and stability, ensuring that you keep delivering great software. To make this possible, it’s vital for technology and operations engineers to collaborate closely. Sharing responsibilities and meeting regularly can help keep everyone on the same page, with a focus on delivery and maintaining smooth operations. This can result in both DevOps and SRE working effectively.