Compute and Billing

Transform pipelines run in isolated ephemeral compute that spins up fresh for each execution and terminates immediately on completion. You are charged only for active execution time - there is no idle or standby cost.

---

What is a compute second?

Compute is measured in seconds of wall-clock execution time. Every second your pipeline is actively running - extracting data, applying operations, loading results - consumes one compute second from your balance.

There is no per-operation surcharge. A Join and a Filter cost the same per second. The total cost of a run is determined entirely by how long it takes.

---

How compute is metered

Compute consumption is tracked phase by phase throughout each run:

Each phase is timed separately. The total compute consumed for a run is the sum of all phase durations.

---

Balance validation before every run

Before a run is authorised - whether scheduled or manually triggered - Edilitics checks your compute balance against the projected cost of the run.

How the projection works:

• New pipelines (no prior successful runs): the projected cost defaults to a conservative baseline.
• Established pipelines: the projected cost is based on the historical average of recent successful run durations.

If your balance is below the projected cost, the run is blocked before it starts. You will see a clear error in the Run History with a link to top up your compute balance.

---

What happens when compute runs out mid-run

If compute is exhausted while a run is in progress, the pipeline terminates immediately and the run is marked Failed. Edilitics does not pause or checkpoint mid-run - the run stops at whichever phase was executing when the balance was exhausted.

The failure is recorded in run history automatically, including the phase at which the run stopped.

---

Resuming after low compute

When resuming a paused pipeline, Edilitics checks that your balance covers at least one projected run before allowing resume. If the balance is insufficient, resume is blocked with a prompt to top up.

---

No data is stored in Edilitics infrastructure

Each run executes transiently. Source data is pulled into isolated memory, operations are applied, and the output is written to your destination database. Edilitics retains only the pipeline configuration - the operation sequence, column mappings, and schedule. No customer records are stored in Edilitics infrastructure at any point.

---

Viewing your compute balance

Your current compute balance and usage history are available in the Compute dashboard, accessible from the main account settings.

The run history for each pipeline shows the compute seconds consumed per run, so you can track which pipelines are the heaviest consumers.

---

Plan compute limits

Each plan includes a monthly base compute allocation, a machine tier that determines processing speed, and a maximum duration per individual pipeline run.

The machine tier also determines row throughput: a Pinnacle run processes significantly more rows per compute second than a Launch run on the same data, because it has more CPU cores and memory available.

---

Addon compute packs

Each plan includes a monthly base compute allocation. When that is exhausted, you can top up with addon compute packs. Packs are one-time purchases and do not expire at the end of the billing period.

Five pack sizes are available, with volume discounts on larger packs:

Row capacity estimates below assume an average row size of approximately 2 KB and 20-40 columns per table - consistent with the column limits per plan (Launch: 50 columns, Scale: 100 columns, Pinnacle: 200 columns). Actual capacity depends on your data's density, the number of operations in your pipeline, and whether operations expand or reduce the column count.

---

Frequently Asked Questions

---

Next steps

Need help? Email support@edilitics.com with your workspace, job ID, and context. We reply within one business day.