Skip to content

SolutionsWeb apps and APIs

Any app, asleep until it is called

Any container image at its own address. It sleeps 100 ms after its last request and wakes in 3.5 ms, so idle hours use no CPU.

Read the Cube docs

Shell
# any image; waits until it is ready, then prints its address
glac deploy ghcr.io/acme/api:1.4 --name api --port 8080
glac cube versions api       # every deploy is a version
glac cube rollback api       # back to the version before
glac cube logs api -f        # its log, a line per request
glac compose up              # a Cube for each Compose service

Paid while nobody calls

A server kept running is paid for every hour of the month. As a Cube, an instance wakes for a request and sleeps 100 ms after the last, its CPU billed by the busy millisecond.

those requests use in a month1
200M CPU-ms
a vCPU kept running all month is paid for1
2.6B CPU-ms

Versions, and instances that wake

Each deploy is a version whose instances start from its template, your service already running. A request wakes one in its region; it sleeps 100 ms after the last, and is deleted after an hour unused.

An app: its versions, the ready one’s template, and its instances in each region, awake for a request and asleep after it.

What your app gets

  • One command: glac deploy prints the address once the version is ready.

  • Wakes with its state: caches, pools and sessions in memory are still there.

  • Versions and rollback: the traffic moves when ready, and back in one call.

  • Scale per region: instances start as requests come, up to your limit.

  • HTTP/3, gRPC and WebSockets, at your own domain with its certificate.

  • Compose, workers and jobs: a Cube for each service, and a job runs to its end.

Read the docs

Illustrative previews

Measured, not promised

Medians on bare metal with nothing else running: a best case, and we say so. A Cube’s instance is the same kind as a sandbox, where these were measured. How we measure

to wake a sleeping instance2
3.5 ms
for the call that woke it, sent to answered2
6.7 ms
to put it to sleep, with its memory and processes2
2.3 ms

See it for your own app once the service opens: the log’s line for a request that woke an instance says how long the wake took.

Shell
glac cube logs api -f
# one line a request; one that woke an instance says how long it took

The products it uses

A month, worked out

The pricing estimate’s first app, at the list prices. With an instance kept awake all month, it would come to $6.84. Estimate your own

The use

  • 10 million requests a month, arriving at random, in one region.
  • 20 ms of CPU each, 256 MiB of memory in use, 10 KB an answer.
  • An instance awake while it answers and for 100 ms after, of 2 vCPUs, the size an app gets unless you say.

A month at the list prices

$5.93

  • CPU200 million CPU-ms × $0.01 / M$2.00
  • Memory66.7 GiB-h × $0.008$0.53
  • Requests10 million × $0.30 / M$3.00
  • Traffic out100 GB × $0.004$0.40

On Paid this month is $6.40: the $5 fee, and $1.40 past what its inclusions cover, from your balance at the same prices.

Deploy the first one

A key, then one command with any image. It prints the address when the version is ready.

CubeThe command line