Hot Standby

Hot standby keeps your JuliaHub environment ready to go so users don't have to wait for compute resources to start up. Without hot standby, launching a session can take several minutes while cloud infrastructure provisions a new machine and downloads the required software. With hot standby, a machine is already running and the software is already downloaded — so sessions start in seconds instead of minutes.

How It Works

Hot standby has two parts that work together:

1. Standby nodes — JuliaHub keeps one or more machines running and ready in the background. When a user starts a session, they get one of these pre-warmed machines instantly, and JuliaHub provisions a replacement in the background.

2. Image caching — Application images (e.g., Julia IDE, Pluto) are pre-downloaded onto the standby nodes. This eliminates the image download step, which can otherwise add minutes to startup.

Both parts need to be configured for the best experience. Standby nodes without image caching still save time (no machine boot wait), and image caching without standby nodes still helps (no download wait on new machines), but together they provide near-instant startup.

Scheduling Standby Nodes

Standby nodes cost money while they're running — even when no one is using them. To balance cost and responsiveness, JuliaHub supports scheduled hot standby. This lets you define when standby nodes should be available, typically matching your team's working hours.

Common Schedules

Here are some examples of schedules that customers commonly use:

Schedules are timezone-aware — if you request "8 AM – 5 PM Eastern," it will automatically adjust for daylight saving time changes.

Outside Scheduled Hours

When hot standby is not active (e.g., evenings and weekends), sessions will still work — they just take the normal amount of time to start up (typically 3–8 minutes depending on the machine type and image size).

Multiple Standby Nodes

You can run more than one standby node at a time. This is useful if many users tend to start sessions around the same time (e.g., at the start of the workday). The first user gets the first standby node instantly, but without a second standby node, the second user would have to wait for a new machine. Running 2–3 standby nodes during peak hours and 1 during quieter times gives a good balance of responsiveness and cost.

Enabling Image Caching

Image caching is controlled per application version in the Application Version Management panel. To enable it:

1. Go to Admin → Application Version Management
2. Select the application (e.g., Julia IDE)
3. Select the repository (CPU or GPU)
4. Toggle Hot Standby on for the image versions you want cached

Cost Considerations

Hot standby nodes use the same machine types as regular sessions. The cost of running a standby node is the same as running an idle session of that machine type.

To estimate the cost impact:

• 1 standby node, business hours only (9 hours/day, weekdays) = ~45 node-hours per week
• 1 standby node, 24/7 = 168 node-hours per week
• 2 standby nodes, business hours only = ~90 node-hours per week

JuliaHub support can help you determine the right machine type and schedule for your team's needs.

Requesting Changes

Hot standby scheduling is configured by JuliaHub support. To request changes:

1. Decide on a schedule — When do your users need fast startup? (See Common Schedules above)
2. Decide on capacity — How many standby nodes do you need during peak vs. off-peak hours?
3. Specify your timezone — e.g., "US Eastern," "Central European Time," "Japan Standard Time"
4. Contact JuliaHub support with your requirements