Pharos Production / Capacity planning

Will your queue
actually catch up?

Estimate recovery time while new work keeps arriving. Compare your successful completion rate with the capacity needed to clear a backlog inside a chosen window.

A WORKING REFERENCE / 001

A faster consumer can still miss the deadline. Start with the net rate, then inspect the work left at the end.

Steady-rate model · No queue connection · Local JSON export

Preparing a wider project?

Browse the review library →

01 / Set the assumptions

Queue drain calculator

Use the same work unit for every rate
Whole work items waiting now.
Continuing incoming work.
Sustained unique completions, not retry attempts.
Time available to reach zero backlog.

Entries stay in this page. Download a draft to keep a local copy. Reloading the page clears your changes.

02 / Read the projection

Enter your assumptions to calculate.

Net drain rate
Pending items/s
Time to zero backlog / minutes
Pending
Completions needed for the window
Pending items/s
Backlog at the deadline
Pending items
Projected backlog over the recovery windowCalculated from constant arrival and completion rates. Exact checkpoints appear in the table below.StartDeadline
Line is a steady-rate projection. It has no measured queue data.
Projection checkpoints
Elapsed minutesWork items
PendingPending

This is a planning estimate. It does not account for bursts, work-item cost, ordering constraints or downstream saturation.

03 / MODEL

Count new arrivals
before dividing.

Subtract the arrival rate from successful completions to find the net drain rate. Divide the initial backlog by that net rate only when it is positive. If completions equal arrivals, an existing backlog stays level. If arrivals are higher, it grows.

net = completions − arrivals
drain seconds = backlog ÷ net, when net > 0
required completions = arrivals + backlog ÷ window seconds
remaining = max(0, backlog − net × window seconds)

The required rate is an average under these assumptions. Add operational headroom through a separate capacity review. The required completion rate is displayed rounded upward to two decimal places. Other display values round to the nearest two decimals; exports retain the unrounded values.

04 / WORKED EXAMPLE

20,000 items.
New work still arriving.

With 25 arrivals and 40 successful completions each second, the queue loses 15 items per second. The initial 20,000 items drain in about 22.22 minutes. A 30-minute window needs about 36.12 successful completions per second when rounded upward to two decimal places.

If successful completions fall to 20 per second, the same backlog grows to 29,000 items after 30 minutes. A queue that is empty at the start can also grow when arrivals exceed completions.

05 / BEFORE USING THE RESULT

Verify the rates
in your own system.

  1. Measure arrival and successful completion rates over comparable windows. Count the same kind of work item.
  2. Separate retries from unique successful work. Include retry costs in the sustained completion rate you can actually deliver.
  3. Check oldest-item age and per-tenant behavior. A small total backlog can hide work that is already too late.
  4. Recalculate when rates change. Validate throughput and downstream limits before changing production capacity.

Read the assumptions, input limits and export fields ↗

REFERENCE / SOURCES

Source notes.

Read alongside the method. External guidance provides context; the tool rules are defined on this site.

Sources reviewed 9 November 2026. Examples below are fictional planning inputs.