Availability vs Tier III
Placing the burner's 0.86-0.995 availability beside the 0.99982 Tier III standard shows exactly how large the gap is and why it drives the architecture.
Reading the nines
Reliability is spoken in 'nines'. Tier III's 0.99982 is about three-and-three-quarter nines: roughly 1.6 hours of downtime per year. Tier IV pushes further still. The burner's high end, 0.995, is between two and three nines — about 1.8 days per year. Its low end, 0.86, is not even a single nine of the fractional part; it is about 49 days per year.
The gap is easiest to feel in downtime rather than in decimals. A Tier III data center expects to be dark for the length of a coffee break across a whole year. A burner at its modelled best would be dark for the length of a long weekend, and at its modelled worst for the length of a season's worth of Fridays. No hyperscale customer accepts either as a sole supply.
The comparison is not meant to discourage; it is meant to size the redundancy correctly. If a single burner delivers 0.995, then reaching 0.99982 requires additional independent capacity whose combined outage probability multiplies down to the Tier III floor. That is standard reliability engineering — the same math that lets data centers hit Tier III today using generators and utility feeds that are individually far less reliable than 0.99982.
The takeaway is quantitative and honest: the burner is a strong member of a Tier III system, contributing firm base capacity, but the Tier III guarantee is a property of the whole hybrid, computed deliberately, not a property inherited from any one machine.
- Tier III = ~1.6 h/yr; burner best = ~1.8 d/yr; worst = ~49 d/yr
- 'Nines' translate to intuitive downtime windows
- Combined systems can exceed any single member's availability
- Redundancy is sized from this gap, not hand-waved