Discover how Edgegap works

Discover how Edgegap works

Idle Machines

Idle machines are servers you are paying for that run no matches, or run far fewer than they could hold. They tend to appear when capacity is reserved ahead of demand: overnight, between peaks, or in regions that never filled. Idle capacity can be one of the largest sources of waste in reserved hosting, and a common reason studios pair reserved machines with on-demand capacity.

Also called

idle servers, idle capacity, idle game servers

By

By

By

Gabriel Parent

Gabriel Parent

Gabriel Parent

,

Director

Director

Director

Published

Published

Published

Last reviewed

Last reviewed

Last reviewed

What Makes a Game Server Machine Idle?

A machine is idle when you pay for capacity that no match is using. That happens in two ways:

  • Fully idle: a host or virtual machine running no match at all. It is easy to spot and, on most setups, short-lived: a scheduler fills it with the next match or releases it. On a monthly reservation it is still billed until the term ends.

  • Partly idle: a machine running matches, with CPU and memory that none of them use. It looks busy on a dashboard, so it can run like this all day, and it shows up mainly on the bill.

The second case is usually the larger cost. In fleets that scale by whole machines, it can be the normal state rather than the exception.

Two other uses of the word sit close by. Empty player seats inside a running match are measured by session fill rate, not machine capacity. A player who stops moving is “idle” too, in the AFK sense, which is a game design question covered in the FAQ below.

Article's key insights

  • What Private Fleets are: reserved, private capacity for dedicated game server hosting on hardware set aside for your studio in a specific location, with no egress fees and no reason to spin servers down between matches.

  • Two sides of one axis: everything below the zero line is your own fleet, everything above it is capacity that spilled over to public cloud.

  • Every bar is an hourly average, not a peak: a 4 vCPU deployment that ran 15 minutes shows up as 1 vCPU used, which is why bursty workloads look quieter than they felt.

  • Idle green is not always usable green: the idle number sums every host in a city, so a fleet with 3 vCPU free across three hosts still cannot fit a 1.5 vCPU deployment.

  • Small deployments can round into invisibility: a 0.25 vCPU server running 6 minutes is 0.03 vCPU over the hour, roughly 0.25% of a 12 vCPU fleet. Hover to see the real value.

  • The last three hours are intentionally hidden so deployments still resolving into an Error state do not inflate your usage.

Where Does Idle Capacity Come From?

Orchestration that scales by whole virtual machines usually runs a set number of server processes on each one. The waste it can create comes in two kinds:

  • Wasted capacity: machines running ahead of or behind players. Where a fleet scales, it adds a whole machine at a time, so each step usually adds more room than the next match needs. Where it doesn’t, it is sized for each location’s peak, and player demand typically peaks in the local evening and falls overnight. Scaling down waits for the last match on a machine to end, which protects live players but keeps a mostly empty machine billed until then.

  • Wasted usage: capacity inside a running machine that no match uses. On a machine you rent whole, part of the CPU and memory you pay for goes to its operating system and agents rather than to matches. Server slots are often sized for the heaviest match, so a lighter match strands the rest of its slot, and where slots are set in whole vCPUs, a match that needs a fraction of one still reserves the whole.

Two charts of one location over 24 hours. Left, whole-VM scaling: virtual machines added one at a time ahead of player demand, each stopping only when its last match ends, with hatched wasted capacity between demand and the machines and a shaded band of unused room inside each machine. Right, capacity in sync with players: a reserved fleet covers a chosen share of traffic, idle only in the quietest hours, and on-demand capacity follows the demand curve above it with a thin band of unused room. Takeaway: machines with no match at all are easy to spot; the common cost is machines that run with more than their matches use.

A bill counted in machine-hours doesn’t separate the two from actual play. Both are paid at the full rate, wherever the machines run.

Ending machines sooner after a peak is covered under auto-scaling, and capacity held ready on purpose, for faster starts, under warm pools.

How Do You Find Idle Capacity in a Server Fleet?

Idle capacity hides in averages. Three habits make it visible:

  • Read it by location and by hour: A fleet that looks reasonably used over a day can be mostly empty before dawn and full in the evening, in the same city. Each location has its own curve, shifted by its time zone.

  • Count capacity, not machines: A host running one small match is not idle by a machine count, but most of its vCPU can be. Compare the vCPU you pay for with the vCPU allocated to running matches.

  • Check whether free capacity can be used: Free capacity spread across several hosts can be too small, on any one host, to fit the next server. That capacity is paid for and idle, while new matches still need room elsewhere.

Reserved capacity adds time: a fleet sized for a launch peak stays partly idle until its term ends, so good fleet management commits only as far as the data supports.

How Can Orchestration Keep Capacity in Sync With Players?

Idle capacity shrinks when capacity is added and removed in the same units players use: one match at a time.

  • A server per match, started when the match forms: a container starts for the match and stops when it ends, so there is no machine to keep warm between matches.

  • Sized to the match: fractional vCPU sizing gives smaller, right-sized dedicated servers, and for very small matches, a multi-room server can hold several in one process.

  • Side by side on one host: many matches share a host, each with its own allocated CPU and memory, isolated by design, so a host fills with matches instead of empty slots (noisy neighbor covers how that isolation works).

  • A reserved fleet for a cost-effective share of overall traffic, and on-demand capacity for everything above it: reserved hosts carry the hours they stay busy, and on-demand servers start and stop with the matches above that. That split is hybrid orchestration.

On Edgegap, Edge Cloud is billed by the minute, only on usage, with deployments from 0.25 to 4 vCPU: with no active servers, Edge Cloud costs $0. Private Fleets are reserved hosts, billed monthly on 1-, 6- or 12-month terms. Matches pack onto them by vCPU, and automatic overflow to Edge Cloud means a fleet doesn’t have to be sized for the peak. The pricing page and its calculator put numbers on both for your game.

A word from our sponsor (ourselves!)

Private Fleets give you dedicated bare-metal compute in 58 locations for persistent, cost-effective, egress-free hosting. When your player base spikes, Edge Cloud automatically absorbs the overflow on demand, keeping online performance steady wherever your players are.

Edgegap's Take (just our opinion, take it with a grain of salt!)

Measure Idle Capacity by Location and Hour, Then Size Your Fleet to the Hours It Stays Busy

Fully empty machines tend to get noticed and shut down. The idle capacity that lasts is the kind that looks busy: a city running well under half its reserved capacity through the night, a host with room no server can fit into, a reservation sized for launch-week traffic that never returned.

A useful weekly check, per location: billed vCPU against vCPU allocated to running matches, hour by hour, over the last 30 days. Our Private Fleet utilization guide reads it the same way: the 30-day view for sizing, the 7-day view for behavior. Capacity idle for most of the day, every day, is reservation to release when the term allows; idle only in the quietest hours is the expected cost of a reservation. Overflow at the same hours every day may be worth reserving, once those hours pass the break-even point in the pricing calculator. Overflow scattered at random is on-demand capacity doing its job.

Before adding hosts, check that free capacity can actually fit your servers: free vCPU split across hosts can still be too small for one deployment.

The aim is a hybrid setup: a reserved fleet for a cost-effective share of overall traffic, and on-demand capacity for everything above it, with auto-scaling that starts and stops servers with each match.

-

-

-

Gabriel Parent

Gabriel Parent

Gabriel Parent

,

Director

Director

Director

Frequently Asked Questions

How do you calculate idle time for a machine?

Do idle servers still cost money and power?

What happens to idle players on a game server?

What does idle mean on a server?

Get your Game Online Easily & in Minutes

Start Integrating Now!

Get your Game Online Easily
& in Minutes