0% completed
DevOps, SRE, Platform, and Production Engineers
On This Page
One Problem, Four Responses
DevOps Engineer
Site Reliability Engineer
Platform Engineer
Production Engineer
Comparing the Roles
Why the Titles Overlap
Using This in an Interview
Key Takeaways
Practice Questions
"Checkout became slow after the last release. Who should fix it?"
At the grocery company, the delay is easy to measure. The checkout service's 95th percentile latency (p95) rose from 400 milliseconds (ms) to 1,600 ms. Latency is the time that a request takes, and p95 is the time within which 95 percent of requests finish. Before the release, 95 percent finished within 400 ms, but afterward 5 percent took longer than 1,600 ms.
A DevOps engineer, a site reliability engineer, a platform engineer, and a production engineer could all help. Each would begin with a different question, and that difference explains the four titles.
One Problem, Four Responses
The release reached all customers during the busiest period and stayed there for 12 minutes before the team reversed it. The four roles respond to different parts of that event.
DevOps Engineer
A DevOps engineer asks why the delivery path allowed this change to reach everyone. The canary compared only the error rate, which is the share of requests that fail. The error rate stayed at 0.2 percent, so the slow version passed.
The engineer adds a latency rule to the pipeline. A canary now fails if its p95 is more than 20 percent above the current version. At a current p95 of 400 ms, any canary above 480 ms is stopped, so this release, at 1,600 ms, would not have passed.
Site Reliability Engineer
A site reliability engineer (SRE) uses software engineering and measured targets to keep a running service reliable. Google created the discipline and published much of its early practice.
An SRE asks how close this event brought the service to missing its reliability target. A service level objective (SLO) is such a target. Checkout's SLO says that 99.9 percent of checkouts must succeed within 1 second over 30 days, so a checkout that succeeds slowly still counts as bad.
At 30,000 orders a day, 30 days contain 900,000 checkouts. The remaining 0.1 percent allows 900 bad checkouts, and this allowance is the error budget. The p95 alone cannot show how many checkouts were too slow, so the SRE counts them. The new query ran on every checkout, so nearly all of them took longer than 1 second during the 12 slow minutes. At the peak rate of 50 checkouts a minute, that is about 600 bad checkouts.
One release used two-thirds of the 30-day budget, so the SRE asks the team to delay risky checkout changes and adds an alert that warns when the budget is being used quickly.
Platform Engineer
A platform engineer builds internal products that other engineers use to build and run software. Those engineers are the platform's customers, so ease of use matters as much as technical correctness.
The platform engineer asks how all 14 services can receive the new rule without extra work. The latency rule moves into the shared deployment template, which is the standard configuration that each service uses. Every team now receives the check by default and can change the limit only through a reviewed setting.
Production Engineer
A production engineer combines software development and systems knowledge to keep specific services running well. Meta is widely known for this title, and its production engineers work closely with service teams.
The production engineer asks why checkout itself became slow. The database's slow query log, which records queries that exceed a time limit, points to a new query. The query has no index, the lookup structure that lets a database find rows without reading the whole table. The team adds the index and releases the change again, and p95 stays near 400 ms.
Comparing the Roles
| DevOps engineer | SRE | Platform engineer | Production engineer | |
|---|---|---|---|---|
| Main question | How does change reach production safely? | How reliable must this service be? | What should every team receive by default? | Why is this service behaving this way? |
| Typical work | Pipelines, infrastructure, automation | SLOs, incidents, reliability engineering | Internal tools and self-service systems | Service performance and operations |
| Main users | Development teams | Service owners and customers | Internal engineers | Service teams |
| Common measure | Delivery speed and safety | SLOs and incident results | Adoption and developer time saved | Service health and efficiency |
A cloud engineer is a related title that usually focuses on designing and operating resources in a provider such as AWS, Azure, or Google Cloud.
Why the Titles Overlap
These titles describe areas of emphasis, not fixed boundaries. At the grocery company, one DevOps engineer performs parts of all four jobs. A very large company may have separate teams for each one, and another company may give the same work a different title.
For that reason, the job description tells you more than the title does. Many postings mention CI/CD, which means continuous integration and continuous delivery: automatically testing each change and preparing or releasing it to production.
| If the posting says | The work is closest to |
|---|---|
| Build CI/CD pipelines and manage Terraform | DevOps |
| Own SLOs, join on-call, lead incident reviews | SRE |
| Build self-service tools for developers | Platform engineering |
| Debug Linux and improve large production services | Production engineering |
| Design and operate AWS accounts and networks | Cloud engineering |
Using This in an Interview
If you are asked about these roles, avoid a memorized definition. Use one example, then connect it to the job you want.
"If a release slows checkout, a DevOps engineer asks why the pipeline allowed it, and an SRE measures the effect on the SLO. A platform engineer makes the fix the default for every team, and a production engineer finds the slow code or query. In this role, I would expect to focus most on the pipeline and the infrastructure, but I would work with all of those concerns."
Expect two follow-up questions:
- "Is SRE simply DevOps under another name?" They share many practices, but SRE gives more attention to measured reliability targets and incident work.
- "Who should own the service after deployment?" The development team should usually own its behavior, while shared teams provide the tools and guidance that make this practical.
Key Takeaways
- DevOps engineers focus on how change reaches production safely.
- SREs manage reliability through measured targets such as SLOs and error budgets.
- Platform engineers turn common needs into internal products.
- Production engineers study the behavior of specific services in detail.
- Job descriptions show the real work more clearly than titles do.
The next lesson, How a Code Change Reaches Production, follows one change through the path that these engineers maintain.
Practice Questions
1. Why did the slow release pass the first canary check?
<details><summary>Show answer</summary>Because the check measured failures but not speed. The error rate stayed at 0.2 percent, so the canary appeared healthy. A release can harm users without increasing the number of failed requests.
</details>2. An SLO allows 900 bad checkouts in 30 days, and 300 have already occurred. How much of the error budget remains?
<details><summary>Show answer</summary>600 bad checkouts, or two-thirds of the budget, remain. The team can use that remaining allowance to judge whether a risky release is acceptable.
</details>3. A posting is titled "DevOps Engineer," but it asks you to own SLOs and lead incident reviews. How should you prepare?
<details><summary>Show answer</summary>Prepare for the work described, which is close to SRE. Keep your delivery knowledge ready, but add practice on reliability targets, on-call decisions, and incident communication. Ask the recruiter which topics the interviews will cover.
</details>Reading Progress
0%
On This Page
One Problem, Four Responses
DevOps Engineer
Site Reliability Engineer
Platform Engineer
Production Engineer
Comparing the Roles
Why the Titles Overlap
Using This in an Interview
Key Takeaways
Practice Questions