Cost Model

What it does

ixtract estimates the cost of each candidate plan before execution, then shows you the trade-off. You declare your rates; ixtract applies them to the planned workers, estimated duration, and estimated egress.

Cost is an estimate, not a guarantee. All rates default to zero. If you don’t declare rates, no cost is shown.

How cost is calculated

cost = (workers × duration_estimate) × compute_rate
     + (estimated_output_gb) × egress_rate
     + (workers) × connection_rate

• compute_rate: cost per worker-second (e.g., EC2 hourly ÷ 3600 ÷ vCPUs)
• egress_rate: cost per GB of output data
• connection_rate: cost per connection-second (for managed databases with per-connection pricing)

Duration estimate comes from the estimator (EWMA + context similarity blend). For first runs with no history, it uses the profiler’s latency estimate.

Cost comparison output

ixtract generates candidate plans at different worker counts, filters dominated candidates, and shows the comparison:

Cost Comparison  (estimated, rates user-declared)
────────────────────────────────────────────────
  workers  duration  cost
  2        ~12s      $0.13  ← planned
  3        ~12s      $0.13
  4        ~13s      $0.15
  8        ~14s      $0.21

The planned option is the one ixtract selected. You can override with --max-workers if you prefer a different trade-off.

A candidate is “dominated” if another candidate is cheaper and faster. Dominated candidates are filtered from the comparison table.

Declaring rates

CLI

ixtract plan orders \
  --compute-rate 0.05 \
  --egress-rate 0.01 \
  --connection-rate 0.001

Or from a file:

ixtract plan orders --cost-file rates.json

rates.json:

{
  "compute_rate": 0.05,
  "egress_rate": 0.01,
  "connection_rate": 0.001
}

Python API

from ixtract import CostConfig, plan

cost = CostConfig(
    compute_rate=0.05,
    egress_rate=0.01,
    connection_rate=0.001,
)

result = plan(intent, cost_config=cost)

if result.cost_estimate:
    print(f"Estimated cost: ${result.cost_estimate.total:.2f}")

CostConfig validates that all rates are non-negative and no unknown fields are present.

How to determine your rates

Compute rate

For a cloud VM running the extraction:

compute_rate = hourly_vm_cost / 3600 / vcpu_count

Example: $0.192/hr VM, 4 vCPUs → compute_rate = 0.192 / 3600 / 4 = 0.0000133

For a managed service with per-hour billing, divide by the total seconds per hour.

Egress rate

Cloud providers typically charge $0.08–$0.12 per GB for data leaving a region. Check your provider’s pricing page.

Example: AWS us-east-1 → egress_rate = 0.09

Connection rate

For managed databases with per-connection pricing (e.g., RDS Proxy, Azure SQL serverless), declare the per-second cost of holding a connection open.

Most setups: connection_rate = 0.

Design principles

• No defaults are ever applied. If you don’t declare a rate, that component of cost is zero.
• Cost does not affect the plan. The planner optimizes for throughput; cost is displayed alongside the plan, not used to select it.
• Rates are user-declared, not system-measured. ixtract does not call any pricing API.
• Comparison uses the same estimator as the plan. Candidates are evaluated with identical assumptions.