What Is an Instance Selection Strategy in Scheduling?

An instance selection strategy is the preference a scheduler applies when several machines inside one work center all have room for an operation: spread the work onto the least busy machine, pack it onto the busiest machine that still fits, or simply take the first machine with any capacity. The strategy never decides whether capacity exists; the capacity arithmetic settles that first. It decides which of several valid answers is chosen. EDGEBIC by User Solutions carries a strategy through a scheduling run and switches it deliberately when scheduling a pinned constraint.

How it works

A work center with more than one machine has to answer "which machine?" every time it books hours. The engine reaches that question only after it has established which machines can genuinely take the work in the slot under consideration, so every candidate already fits. Three preferences then decide between them.

Load balancing picks the least loaded candidate. Work spreads evenly, machine loads stay similar, and an operation with enough hours can be split across several machines to finish sooner. This is the ordinary default and the intuitive answer.

Maximize utilization picks the most loaded candidate that still has room. Work packs tightly, one machine at a time, and machines fill in order rather than in parallel. The result looks worse on a load chart and is often better in practice, for one reason: packing leaves whole machines genuinely clear. Four hours spread across four machines leaves no machine free for a long job; four hours packed onto one leaves three.

First available takes the lowest-numbered candidate with any capacity, ignoring how loaded it is. It is the simplest rule and the least interesting, but it produces stable, predictable placement, which is occasionally exactly what you want.

The strategy is a run-level preference rather than a field on a routing step. What you configure is one level up: the work center's own allocation behavior, which decides whether an operation's hours spread across machines at all, or whether one machine is dedicated to one job for the day. Once that behavior has produced a set of candidate machines, the strategy chooses among them.

One switch is worth knowing about because it is automatic. When the engine schedules a job's pinned constraint operation, it temporarily raises that step's placement priority and switches to packing, then restores both afterwards. The reason is direct: at a constraint, gaps are the enemy, and packing is what fills the resource without leaving unusable slivers between jobs.

A concrete example

A work center runs three machines on an eight hour day shift. Existing bookings have left machine one with six hours used, machine two with two hours used, and machine three empty. A new operation needs two hours on Monday.

Every machine has room, so all three are candidates. The strategy decides.

Read the middle row against the first. Balancing produces loads of six, two, and two: tidy, and with no machine offering a full uninterrupted day. Packing produces eight, two, and zero: lopsided, and with one machine completely open. If the next job to arrive needs a solid eight hour run, only the packed plan can give it one on Monday.

That trade is why packing is the constraint's preference. Suppose this work center were the plant bottleneck and a pinned operation needed those two hours. Packing them onto the machine already at six hours completes that machine's day exactly and leaves the other two intact for the next constraint job. Balancing would have left three machines each partly used, which on a bottleneck is capacity you cannot easily sell.

Notice also that no strategy can overload anything. Machine one was offered the work because it had exactly two hours free, and after the booking it holds eight, which is its shift capacity. Had the operation needed three hours, machine one would not have been a candidate at all and the choice would have been between machines two and three.

How EDGEBIC uses it

The strategy sits at the innermost point of capacity allocation, after the search has found slots with room. The unit it hands work to is described in what is a machine instance in scheduling, and how much one instance can supply is covered in per instance capacity in scheduling.

Two settings sit above the strategy and usually explain an uneven-looking plan before it does. The work center's dedicated-machine behavior is described in the one per day rule in scheduling, and the rule that keeps one job on the machine it started on is covered in what is global instance assignment in scheduling. Between them those two account for most apparent imbalance.

The automatic switch to packing happens on the constraint path, which begins with the pin described in what is an anchor step in scheduling. And the load figures the strategy compares are the same ones reported by what is capacity utilization in scheduling, which is the surface where the consequences of the choice actually become visible.

The takeaway

An instance selection strategy is a tiebreak, not a capacity decision, and reading it that way keeps expectations right: it can never overload a machine, only redistribute hours among machines that already fit. The genuinely useful insight is that even loads are not automatically the goal. Packing looks worse and often serves a plant better, because a clear machine is worth more than a lightly used one, which is exactly why a constraint is packed rather than balanced. For the settings that sit above it see what is global instance assignment in scheduling and the one per day rule in scheduling, then explore EDGEBIC or, if you are coming from the legacy product, RMDB to EDGEBIC.