skip to content

A team calls its system fully managed — why does the ladder apply per component rather than to the system?

level: seniorimportance: should knowfreq 45%

answer

  1. systems do not have a rung
  2. components do, and they differ
  3. the floor is the minimum, not the mean
  4. the apparatus is the cost, not the count
  5. one leftover machine keeps everything alive

basics

~20 s

A system is a set of components and each one sits on its own rung, so fully managed is almost never true of all of them. The routine operational work of the system is set by its lowest-rung component, not by an average.

solid answer

~40 s

Inventory any real system and you find a request-serving service, a data store, a scheduled batch job, a message broker, a telemetry agent, and usually one odd machine running something nobody could package. Those sit on different rungs, and "we are fully managed" is normally true only of the piece the speaker works on. It matters because the operational floor does not average — it is set by the lowest rung present. If one machine still needs patching, the team still needs the whole apparatus for patching a machine: the access path, the runbook, the maintenance window, and a person who remembers all of it. That apparatus costs nearly the same for one machine as for ten, which is why a single hand-operated component can dominate the routine work of an otherwise managed estate.

go deeper

for a junior

Notice that a system is made of several pieces and they are not all run the same way. The data store and the nightly batch job may sit on very different rungs.

for a middle

Explain why fully managed is a claim about one component. Be able to list the work that remains once a single hand-operated machine is left in the estate.

for a senior

Do the inventory out loud: component, rung, hand-performed verbs, runbook owner. Name the component that sets the floor and what it costs the team in routine work.

for a principal

Own the floor as a decision: either fund the apparatus a hand-operated component needs, or spend once to remove the last one — and be able to say which you chose and why.

## A system has no rung; its components do The managed ladder is a property of a thing somebody operates, and a system is not one thing. It is a collection of components, each of which was placed on a rung at a different time, by a different person, for a different reason. So "we are fully managed" is a claim that has to be checked component by component, and it usually turns out to describe the component the speaker personally works on. ## Do the inventory The exercise is dull and it is the whole answer. List the components, write the rung beside each, and beside that the verbs still performed by hand. | component | typical rung | verbs still performed by hand | |---|---|---| | request-serving service | managed runtime | none routine | | primary data store | whole capability | none routine | | scheduled batch job | rented machine | patch, resize, restart, back up | | message broker | whole capability | none routine | | telemetry agent on each host | follows its host | upgraded with the host | | the vendor-supplied binary nobody could package | rented machine | all of them | Two rows in that table keep an entire capability alive, and neither of them is the one anybody presents at a review. ## Why the floor does not average The intuition to break is that a mostly-managed estate is mostly free of machine work. It is not, because the cost is in the **apparatus**, not the instance count. To patch one machine a team needs: - an access path to it, and something that governs who may use that path; - a patch process somebody maintains and occasionally tests; - a maintenance window and a way of telling users about it; - monitoring of the box itself, distinct from monitoring of the service on it; - a runbook, and at least one person who has read it recently. Every item on that list is a fixed cost. One machine needs almost all of it; ten machines need barely more. Worse, an apparatus used by exactly one component is the first to rot: the runbook goes stale, the access path breaks quietly, and nobody notices until the night it is needed. ## The second consequence: what a move costs The lowest rung also prices your next migration, upgrade or recovery. A component that exists as a declaration — an artifact plus configuration, or a managed capability plus its settings — can be recreated somewhere else from that declaration. A hand-built machine has to be rebuilt by whoever remembers what was done to it, and that knowledge is usually undocumented and frequently gone. The hand-operated component is simultaneously the slowest part of any move and the riskiest, which is exactly the combination that makes teams defer it, which makes it worse. ## What a senior actually does about it 1. **List components, not the services you enjoy talking about.** The scheduled job and the agent count. 2. **Write the rung beside each**, then the verbs still done by hand, then who holds the runbook for them. 3. **Look for verbs nobody claims.** The batch machine everyone forgot is where the surprise lives, and it is normally found during an incident or an audit rather than during design. 4. **Price the floor deliberately.** Either fund the apparatus knowingly, or spend once to remove the last hand-operated component and then delete the apparatus with it. Keeping the apparatus by accident is the only option that is always wrong. ## The honest caveat None of this argues that every component belongs on the highest rung. A higher rung costs more per unit, some components have no managed equivalent at all, and a few genuinely need control the upper rungs do not give. The claim is narrower and harder to dodge: **know where your floor is, and pay for it on purpose.** A team that can name its lowest-rung component, the verbs it still performs by hand and the person who performs them is in a completely different position from one that believes the word "managed" applies to a system.

  • Why does one leftover machine cost so much more than its share?
    Because the cost sits in the apparatus, not the count. One machine still needs an access path, a patch process, a maintenance window, monitoring of the box itself, and somebody who knows all of it. Those are fixed costs that do not shrink to one machine's worth, and they rot fastest when only one component uses them.
  • What does the lowest rung do to a migration or an upgrade?
    It sets the price. Components that exist as a declaration can be recreated elsewhere from that declaration; a hand-built machine must be rebuilt by whoever remembers what was done to it, and that knowledge is usually undocumented. The hand-operated component is both the slowest and the riskiest part of any move.
  • Does this mean every component should move to the highest rung?
    No. Higher rungs cost more per unit, some components have no managed equivalent, and a few need control the upper rungs withhold. The point is to know where the floor is and fund it on purpose, rather than discovering it during an incident or an audit.

saying these in an interview costs you the question

  • Says the system is fully managed because its main service is.
  • Averages the rungs of the components into one claim about the system.
  • Assumes one leftover machine costs one machine's worth of operational work.
  • Thinks the lowest rung matters only in proportion to its traffic.
  • Believes moving every component to the highest rung is always right.