Capacity Upgrade Brief Report

Capacity decisions become urgent before a resource reaches 100%. Queueing, latency, failed writes, connection refusals, packet loss, or reduced failover margin can appear while a dashboard still shows unused headroom. The useful question is how much time remains before an operational threshold is crossed.

A credible upgrade brief connects five facts: the current sustained peak, the rate at which demand is changing, the action and critical thresholds, the time needed to deliver usable capacity, and the expected relief from each option. Leaving out any one of them can make a cheap option look safe or a large purchase look necessary when neither conclusion follows from the evidence.

Compounded growth makes timing non-linear. Five percent monthly growth does not add the same number of utilization points every month; each month grows from the previous month's higher level. A flat-growth assumption produces no future breach when the current value is below the threshold, while a small positive rate can eventually cross it.

Capacity gain also needs a precise meaning. A 20% gain means usable capacity is multiplied by 1.20, so the same demand consumes the old utilization divided by 1.20. It does not mean subtracting 20 percentage points from utilization.

Forecasts remain assumptions. Seasonality, a launch, a migration, caching, traffic shaping, or a new bottleneck can change demand after the brief is written. Option cost, lead time, disruption, and claimed gain need owner confirmation before approval.

The aim is not maximum headroom. It is enough validated capacity, delivered before the action window closes, with cost and disruption made visible.

How to Use This Tool:

Use measurements and option estimates that can survive engineering and budget review.

1. Enter the Service or platform, its Current peak utilization, expected Monthly demand growth, and a planning horizon from 1 to 60 whole months.
2. Set an Action threshold and a higher Critical threshold. Correct the thresholds if the critical value is equal to or below the action value.
3. Enter the end-to-end Implementation lead time, choose Match, Lead, or Lag strategy, and set the Target reserve after upgrade.
4. Write concise Bottleneck evidence that names the constrained resource, measurement window, and observed service consequence.
5. Add 1 to 20 upgrade options as comma-separated rows: name, cost, capacity gain percent, lead months, disruption from 1 to 5, and note. Capacity gain must be positive.
6. Set forecast confidence, cost unit, optional budget ceiling, and maximum acceptable disruption when those constraints belong in the decision.
7. Read Approval brief and Option ranking together. Validate the leading option's gain, delivery date, cost, and disruption before using it as a recommendation.

Interpreting Results:

The urgency label compares current and projected utilization with the thresholds and implementation window. Critical now means current utilization is at or above the critical threshold. Start now means the forecast reaches critical or the action-threshold runway is no longer comfortably longer than lead time.

• Approve in horizon means projected utilization reaches the action threshold within the selected horizon.
• Monitor means the selected forecast stays below the action threshold; it is not a guarantee that demand or the bottleneck will stay unchanged.
• A first-ranked option is the highest result from the documented planning heuristic. Budget and disruption penalties are visible, but the rank cannot validate vendor claims or architectural fit.
• If an option says Partial relief only, its capacity gain does not bring horizon utilization below the action threshold.

Technical Details:

The model treats utilization as demand divided by usable capacity and applies a constant monthly growth rate. This is an exponential planning curve, not a regression fitted to telemetry. Every comparison uses full-precision values; display formatting is applied afterward.

Formula Core

Projected utilization after m months is current utilization multiplied by the compounded monthly growth factor.

When current utilization is below a threshold and growth is positive, threshold runway is the logarithmic inverse of the same curve.

Runway is 0 when current utilization is already at or above the threshold. It is reported as no breach at the current trend when current utilization is non-positive or monthly growth is zero and the threshold has not been reached.

The target utilization is the action threshold minus the requested reserve and a strategy adjustment. Lead planning subtracts 5 additional percentage points, Match subtracts 0, and Lag adds 4. The result is limited between the smaller of 50% or one point below the action threshold, and 0.1 point below the action threshold.

An option with gain q% uses a capacity factor of (1 + q/100). Current and projected utilization after that option are divided by this factor.

Rule Core

Option-ranking mechanism

Each option receives weighted contributions for capacity relief, additional runway, relative cost, delivery within the available runway, and lower disruption. The weights are 42, 22, 14, 12, and 10. Lead strategy multiplies the subtotal by 1.08, Match by 1, and Lag by 0.94. An option over the budget ceiling loses 18 points; disruption above tolerance loses 12.

Ties are resolved by lower cost and then higher capacity gain. These weights are a planning heuristic, not an industry standard. A technical or contractual constraint can legitimately override the rank.

Accuracy Notes:

Use comparable telemetry windows and a resource-specific threshold. CPU, storage occupancy, I/O latency, session pools, throughput limits, and queue depth do not fail at the same percentage.

• Recalculate when the measurement window, demand drivers, thresholds, quotes, lead time, or disruption tolerance changes.
• Model a range of growth rates when confidence is low; the selected confidence label changes report wording, not the forecast.
• Validate capacity gain with vendor evidence, load testing, or a controlled rollout.
• Check whether relieving one resource moves the bottleneck to another component.

Worked Examples:

References:

• Architecture strategies for capacity planning, Microsoft Azure Well-Architected Framework.
• Handling Overload, Google Site Reliability Engineering.
• SRE Best Practices for Capacity Management, Google.