Skip to content
New here? Rent a GPU by the second
Kracht
Sign inSign up
Use cases
Computeruns today
Compute

Fine-tune on a schedule

A job that runs nightly needs a GPU for the hour it runs, not for the day around it. Per-second metering makes the difference the whole business case.

A nightly job needs a GPU for the hour it runs. Most billing makes you pay for the day around it.

Why it is hard

  • Hourly rounding eats the job

    A forty minute run billed to the hour is half waste. Billed to the day, it is almost all waste, and the ratio gets worse as the job gets faster.

  • Keeping it warm costs more than the work

    The usual workaround is leaving the machine up between runs, which turns a scheduling problem into a permanent line item.

  • The bill is hard to attribute

    When one machine serves several scheduled jobs, working out which one is expensive means reconstructing it from timestamps.

What answers it

  • Metered by the second

    Forty minutes bills forty minutes. The arithmetic is exact rather than rounded, held in a euro ledger to six decimal places.

  • Nothing runs between runs

    Stop the container and the meter stops with it. There is no warm state to pay for because there is no warm state.

  • Each run is its own line

    Every rental is a separate entry with its own start, stop and rate, so the expensive job is the one with the large number next to it.

story 2 of 20Next: Serve inference when traffic spikes

How billing works

usage_eventsbilled per second, EURlaunchmeter starts with the containerheartbeat10s at the price frozen at matchstopmeter stops, nothing after
heartbeatsent twiceusage_eventsledger_entriescharge · provider_share 85%account_balancesa view: sums, never a columnone fresh row, summedthe retry: unchangedunique(instance_id, seq)

per second, to the microeuro · no minimum · no egress fee · idle costs nothing

Metering starts when the container does and stops when it stops. The rate is frozen the moment you match, so a provider re-pricing cannot change a running rental, and an aborted run costs what it used.

What you use from Kracht

Compute

Rent GPUs by the second from idle hardware, or earn from the cards you own. Priced at match, settled in euro.

Related jobs

Start it from the CLI in your scheduler, and stop it when the job exits.

Start now
Other use cases