The Second Node Buys Uptime, Not Capacity

The Second Node Buys Uptime, Not Capacity

Ada re-derives this chapter’s own numbers step by step, at full precision

foundations
math-foundations
calculation-audit
mqtt
Ada ADA · CALCULATION AUDIT

The Second Node Buys Uptime, Not Capacity

This chapter’s smart building runs 4,000 devices and must hit 99.9% uptime — no more than about 8.76 hours of downtime a year. The chapter is blunt that a single broker handles the traffic easily, yet one node at a typical 99.5% uptime still misses that target, so the design pays for a second node. This audit checks the availability arithmetic to ask what that extra node actually buys, since it adds no capacity the building needs.

Companion to the chapter MQTT Production Operations — every number here comes from that chapter.

See the relationship before changing it

The figure reads from left to right. The blue card is single-node uptime. The middle card applies the page rule. The green card is two-node downtime. Walk the arrows once: set the input, apply the rule, then read the result with its unit.

Single-node uptime changes two-node downtime An input card leads through the rule downtime = (1 - uptime)^2 x 8,766 h x 60 to the two-node downtime result. INPUT PAGE INPUT APPLY THE RULE predict calculate check units OUTPUT RESULT
Walk the arrows. This independence model explains redundancy value, but shared failures remain outside it.

Derive the baseline in four named moves

  1. 1

    Name the input. The chapter baseline is 99.5 %.

  2. 2

    Name the relationship. downtime = (1 - uptime)^2 x 8,766 h x 60

  3. 3

    Substitute with units. (1 - 0.995)^2 x 8,766 x 60 = 13.15 minutes/year

  4. 4

    Read the result. Keep the unit beside the value. Use it only inside the technical boundary on this page.

Predict, then change single-node uptime

Try Predict the direction of downtime = (1 - uptime)^2 x 8,766 h x 60. Test another single-node uptime, then compare two-node downtime.

99.5 %
Chapter baseline
Two-node downtime

Observe This independence model explains redundancy value, but shared failures remain outside it. Reset single-node uptime to 99.5 and compare two-node downtime.

Explain This independence model explains redundancy value, but shared failures remain outside it.

Check yourself

What should you do before trusting a moved-control result?
Answer: Predict its direction, apply the shown relationship, keep the units, and reset to the worked baseline.
What does this small model leave out?
Answer: Only single-node uptime moves here. Field effects named in the technical boundary stay fixed.
TryRecompute Cluster availability: 1 - (1 - 0.995)^2 = 1 - 0.005^2 = 1 - 0.000025 = 0.999975, i.e. 99.9975%.
ObserveTrack Cluster downtime: 0.000025 x 8766 = 0.219 hours = 13.1 minutes/year.
ExplainExplain why two independent 99.5%-available nodes yield 99.9975% cluster availability but add no throughput capacity.

Ready: use the stated baseline inputs, then compare each displayed result.

Ada: Step 2 already showed one broker handles this building’s load, yet the design still specifies two nodes. Let me audit the availability arithmetic that justifies the extra node, because it is the whole reason to pay for redundancy the throughput does not need.

First the target. A 99.9% SLA allows 0.001 x 365.25 x 24 = 0.001 x 8766 = 8.77 hours of downtime per year. A single node at a typical 99.5% uptime is down 0.005 x 8766 = 43.83 hours/year, failing the target by nearly 5x.

Now two independent nodes; the cluster is down only when both are:

  • Cluster availability: 1 - (1 - 0.995)^2 = 1 - 0.005^2 = 1 - 0.000025 = 0.999975, i.e. 99.9975%
  • Cluster downtime: 0.000025 x 8766 = 0.219 hours = 13.1 minutes/year

That is an improvement of 43.83 / 0.219 = 200x (equivalently 0.005 / 0.000025), and it clears the 8.77-hour budget with roughly a 40-fold margin. The point the numbers make is that the second node contributes no throughput the building requires, since Step 2 proved one node suffices, so its entire job is to convert a single point of failure into 200x less downtime. Redundancy here is bought for availability, and the per-device cost of that insurance is $175 / 4000 = $0.044/device/month.

Every number above is taken from the chapter’s own material and re-derived step by step.

Technical boundaries. This availability comparison deliberately does not simulate correlated failures, failover delay, load imbalance, or repair-time distributions. It uses the chapter's fixed node availability assumptions to contrast redundancy with capacity.