Few phrases travel through aviation technology departments as quickly as “we need SRE.” Few are as misunderstood. Site reliability engineering, usually shortened to SRE, is a way of running online services that began in software operations and now shows up in every sector that depends on a website, a booking engine or a training portal, flight academies included. The ideas are sound. The folklore around them is not. Here are six myths worth retiring, each tested against the Ops Atlas description of the discipline.
Myth 1: SRE is just the old operations team with a new badge
Renaming a help desk does not change what it does. The Atlas, which maintains a long entry on SRE, calls this pattern “fake SRE” and lists it as one of the commonest ways for the approach to fail. Real practice rests on measurable user-facing targets, a written policy for what happens when those targets slip, and a cap on repetitive manual work.
An aviation parallel helps. A dispatch office that reprints its rota with a more modern title is still the same dispatch office. A dispatch office that starts publishing departure reliability figures, agrees limits with the chief pilot and acts when those limits are reached has actually changed how it works.
Myth 2: The goal is maximum reliability
Intuition says a safer system is a better one, so more reliability must always be good. In airframes, that logic has a price tag attached, and operators balance it against cost, schedule and weight. Online services work the same way. The Atlas states that reliability is deliberately spent: if a service is running far better than its target, the spare margin is a signal to ship improvements faster, run experiments or retire effort that no longer buys anything.
The source adds a pointed observation about cost. Each additional increment of dependability can cost many times more than the last, while customers stop noticing past a certain level. A student booking a simulator session cares that the page works when they sit down to book; they cannot tell the difference between very good and astonishingly good. Money spent chasing that last sliver is money not spent on the new feature the same student has been asking for.
Myth 3: One hundred percent is a respectable target
It sounds courageous to promise that nothing will ever fail. It is actually a recipe for paralysis. If any failure is a breach, every software change becomes forbidden, since changes are the most common trigger of trouble. The Atlas puts about seven in ten outages down to changes made on live systems, which is why zero tolerance and any kind of progress cannot coexist.
Flight operations never aim for zero in this way either. They aim for defined margins, redundancy and fast recovery. Applied to a website, that means choosing a target such as 99.9 percent over four weeks, noting that it leaves a little over forty minutes of total failure to spend, and agreeing how the spend will be governed.
Myth 4: Blameless means nobody is accountable
Critics hear “blameless postmortem” and imagine a team shrugging after every failure. The Atlas draws a sharper line: blamelessness refers to not punishing people for honest mistakes, while remaining strict about actions, owners and deadlines. A review that never closes its action items is listed among the recognised anti-patterns.
Safety reporting in aviation offers the closest analogy. Open reporting cultures work because people are protected when they disclose errors, and they only improve safety because the findings lead to changes that someone is responsible for completing. Accountability shifts from “who did it” to “who will fix it and by when.”
Myth 5: SRE is the same thing as monitoring or incident response
Dashboards and incident roles matter, but they are neighbours rather than the core. According to the Atlas, the discipline’s own scope covers target design, error budgets, alerting driven by how fast the budget is being used, limits on toil, a postmortem culture, on-call load and capacity planning. Telemetry collection and the choreography of an incident belong to adjacent practices that SRE depends on without absorbing.
Think of a cockpit. Instruments feed the crew information, and emergency procedures tell them what to do in a crisis. Neither one tells an operator how much schedule risk it is willing to accept this quarter. SRE sits closer to that last question, using numbers to decide how fast to move.
Myth 6: Developers never carry a pager once SRE arrives
Some managers hope that hiring reliability engineers means the people who write the software can wash their hands of live service. The Atlas says the opposite. In the originating approach, excessive operational load is pushed back to the developers, and a cap on toil needs real enforcement, such as redirecting tickets or putting developers back on call when the manual burden grows too large.
The healthy arrangement looks like shared ownership. The same principle is familiar on the hangar floor: the people who modify an aircraft’s systems remain answerable for how those changes behave in service. A rota that spreads the load, with each person disturbed rarely enough to stay sharp, is far healthier than one where a single heroic colleague answers every call. The source suggests that a rotation needs around eight people for round-the-clock cover from one site, which is why small teams often share the duty across departments rather than pretending they can cover it alone.
How to spot a myth in your own office
Myths rarely announce themselves. They show up as ordinary sentences in meetings, and three questions are enough to flush them out.
- “What number are we aiming at, and who agreed to it?” If nobody can answer, there is no target, only a hope, and the first myth is probably at work.
- “What happens when we miss it?” An empty answer means there is no written policy, so targets exist only until the first awkward release deadline.
- “Who closed the actions from the last outage?” Silence here points to the blameless-but-toothless version of a review.
None of these questions requires technical vocabulary, which is deliberate. An academy director, a chief instructor and a front-desk manager can all ask them, and the answers say more about operational health than any dashboard colour.
What to do instead
If you are weighing SRE for an academy’s web systems, begin with what students see. Define one or two journeys, such as signing in and booking a lesson, and measure how often they succeed. Decide on a realistic target, work out the allowance and write down the response when it runs low. If you want to see the operational side of the same school, the announcements about an instrument-procedures simulator joining the programme and about a glass-cockpit twin trainer entering the fleet show how much equipment now depends on dependable scheduling systems around it.
The bigger takeaway is that SRE is a set of habits, not a job title. Teams that treat reliability as something they can budget, measure and discuss openly tend to ship with more confidence. Teams that treat it as a vague aspiration tend to alternate between panic and paralysis. Dropping these six myths is the cheapest improvement available. It costs nothing except a willingness to look at the numbers, and the first useful figure is often sitting in a log file nobody has opened.