Skip to Content
🧭

How it works

Updated 2026-11-20

A cost pool is one thing you deliver: a server cluster, a storage system, an application you host for customers, a support team. Everything else hangs off the pool.

Part of a poolWhat it isExample
Recurring CostsWhat the pool costs each month: rent, supplier bills, depreciation of equipment, a share of salaries.Server rental 4,000 per month
Capacity GroupsThe raw physical resource you have, before it is turned into something sellable.640 processor threads, 40 TB of disks
OutputsWhat you actually sell, made from a capacity group.vCPU, GB of memory, GB of storage, a user seat
OccupantsWho uses the outputs, how much, and what they pay.A customer with 8 vCPU and 32 GB
ProductsYour catalogue products, described as quantities of outputs.A plan = 2 vCPU + 4 GB + 80 GB

From these, Lynx works out a unit cost for each output, a cost for each product, the live margin of the pool and how much room is left before it is full.

Sellable capacity

Not all of a raw capacity can be sold. Lynx starts from the Raw Capacity of the group, takes away what is Reserved (for the platform itself, or kept free to survive the loss of one machine), and converts the rest into units of each output.

  • Overcommit: one processor thread can be sold as several vCPU. An overcommit of 4 turns 100 threads into 400 vCPU; 1 means none.
  • Replicated storage keeps several copies of each unit: with 3 copies, 30 TB of disks gives 10 TB to sell.
  • Erasure-coded storage splits each unit into data parts (k) and parity parts (m): with 4 + 2, 30 TB of disks gives 20 TB to sell.
  • Yield after loss: for anything where part of the capacity is always lost, such as staff hours that are not billable.

Tip: Several outputs can share one capacity group, for example two storage types on the same disks. The free space of the group is counted once, so each output shows the room really left and not the same space twice.

Was this article helpful?
Thank you for your feedback. Still stuck? Contact us