What Is Per-Instance Capacity in Scheduling?

Per-instance capacity is one machine's share of a work center's capacity in a given shift and date, obtained by dividing the work center's shift capacity by its number of machines. It is the ceiling the scheduler applies when deciding whether a specific machine can take on more hours. EDGEBIC by User Solutions computes the work center's figure once and derives each machine's share from it, so every adjustment applied at the work center level flows through automatically.

How it works

A work center's capacity for one shift on one date is computed from a single formula. The shift's duration has downtime and any partial-holiday overlap subtracted, then the remainder is multiplied by the machine count and by the work center's utilization percentage. That produces one number for the whole resource, and it is stored on the capacity slot for that combination of work center, shift, and date.

Per-instance capacity is that number divided by the machine count. If a work center has three machines and its shift capacity is twenty-four hours, each machine's share is eight.

The share matters when the scheduler allocates. Before placing hours on a machine, it checks how much that machine has already consumed and compares it against the share. If the remaining room is effectively zero, the machine is skipped and the next one is considered. This is what keeps a single machine from being booked beyond its physical day even when its neighbors are idle, which would otherwise be an easy mistake for a pooled capacity number to allow.

How hours are then distributed across the machines depends on the work center's settings. Load balancing spreads the hours evenly so the machines finish together. Sequential filling packs one machine before moving to the next. A one-job-per-day rule dedicates a machine to a single job for the whole day regardless of how few hours it uses. In all three cases the per-instance share is still the ceiling; only the distribution differs.

Machine counts can be fractional, which is how you model a unit that is only partly available or runs at reduced rate. The total capacity reflects the fraction and so does each share, which is more honest than declaring a whole machine and then trying to hold half of it back.

A concrete example

Think of a taxi rank rather than a taxi company. The rank has three cars and each driver works an eight-hour day, so the rank offers twenty-four driver-hours. That total is a useful planning number, but you cannot hand one passenger a twelve-hour ride and expect a single car to deliver it inside a shift. One car owns eight of those hours and no more.

Now put a job on the rank. A four-hour job fits on any car. A twelve-hour job does not fit on one car at all, and the dispatcher has two choices: split it across cars if the work can be shared, or let it run past the end of the shift and finish tomorrow on the same car.

Change one thing. A car goes in for a two-hour service that morning. The rank's total drops accordingly and so does that car's share. Nobody has to re-enter anything per car, because the service was recorded once against the rank and the shares are derived from the total.

That is exactly how a work center and its machines behave. One capacity number for the resource, divided into shares, and every adjustment applied once.

How EDGEBIC uses it

The share is derived inside the allocator on every scheduling run, so it is always current with the work center's configuration. The individual machine it applies to has its own glossary entry in machine instance, and the remaining room on the whole resource is covered in available capacity.

How the total is derived, including where downtime and partial holidays are subtracted relative to the multipliers, is set out in how capacity is computed for a work center, shift, and day. The multiplier that most often surprises people has its own entry in work center utilization percentage.

Which machine actually receives the hours, and why a job tends to stay on the machine it started on, is walked through in how EDGEBIC picks an instance: load balancing vs one per day. Setting the count and the percentage together is covered in how to configure instances and utilization in EDGEBIC. For the wider vocabulary, see the manufacturing glossary, and to see capacity shares inside a live plan, explore EDGEBIC.