Meter
The running count of hours for which compute is held for a job. It starts when compute is held, stops when the job finishes, and is the only thing a team pays for.
· 2 min read
IN SHORT
The meter is the count of hours compute is held for a job. In E31 Network's design it starts when compute is held, runs while the job executes, and stops when the job finishes and the compute is released. It does not run while a job is queued. A team would pay for what the meter counted at a published hourly rate; no rates are published yet.
The meter is the running count of hours for which compute is held for a job. It starts when compute is held and stops when the job finishes and the compute is released. Everything a team pays for on E31 Network is what the meter counted.
Why it matters
A meter makes the cost of a job legible. The team can see, for each job, how long compute was held, and the bill is that duration at a published hourly rate. There is no separate line for access, no charge for time spent waiting in a queue, and nothing to pay after the compute is released.
That legibility changes how teams think about work. When the meter is per job, a team can ask what a retraining run or a footage pass actually costs, compare it to owning the hardware, and decide job by job. Owned capacity hides that cost inside a fixed line; a meter puts it next to the output it produced.
How it works on E31 Network
In the life of a job, the meter runs through two of the five steps. It starts at rent, when compute is held for the job, and it runs through run, while the job executes. It does not run during submit, while the job is described. It does not run while the job sits in a queue waiting for compute. It stops at release, when the output has been returned and the compute goes back to the pool.
The unit is the hour. A team would pay for the hours compute is held for its job, and the meter stops when the job finishes. The design intent is that what was held is accounted for after each job, so a team can see what it used in a form someone outside the team could read.
No rates are published today. E31 Network is a technical preview, and rates would be published when access opens. Nothing about minimums, discounts, credits, or commitments has been published either, and nothing beyond "by the hour, meter stops when the job finishes" should be assumed. See how it works for where the meter sits in the life of a job.
Related terms
Questions this raises
- Does the meter run while my job waits in a queue?
- No. The meter starts when compute is held for the job, which happens after the wait ends. Queue time is not metered, so a job that waits for the afternoon and runs tonight is billed only for tonight.
- Is there a minimum charge or a commitment?
- Nothing of the kind has been published. The published design is by the hour, with the meter stopping when the job finishes. Rates and any surrounding terms would be published when access opens, and none should be assumed before then.
- What record does the meter leave?
- The design intent is that what was held for each job is accounted for after the job, in a form someone outside the team could read. The exact format is not yet published.
KEEP READING