Spot servers in Spinup let you take advantage of unused AWS capacity at deeply discounted rates—often up to 70% cheaper than regular instances. They are ideal for short-lived, non-critical workloads that can tolerate interruptions.

What Are Spot Servers?

Spinup Spot servers are built on AWS Spot Instances and managed using Spot by NetApp (formerly SpotInst). These servers are automatically restarted if interrupted, and retain their IP address and data volume.

Key Limitations

Before using Spot servers, keep in mind:

  • You cannot choose a specific instance type—only a general size (e.g., spot-small)

  • No cost estimate is provided at launch; pricing fluctuates with AWS demand

  • No shutdown, resize, or snapshot support

  • No image creation or disk modifications

  • No additional disks or disk resizing

  • May be preemptively stopped by AWS at any time

⚠️ Spot servers are automatically recreated if stopped by AWS, but manual shutdowns from the OS (e.g., shutdown -h now) will also trigger this behavior.

Recommended Use Cases

  • Dev/test environments

  • Short-lived compute jobs

  • Cost-sensitive workloads that can restart if interrupted

How to Launch a Spot Server

  1. Log in to Spinup
    Visit https://spinup.internal.yale.edu and sign in with your Yale NetID.

  2. Create a New Server
    In your Space, click Add Server.

  3. Select an Operating System
    Spot works with most OS options (Linux, Windows, etc.).

  4. Choose a Spot Instance Size
    Under Instance Type, select a spot- prefixed size (e.g., spot-micro, spot-small, spot-large).

  5. Review the Allocation Order
    Each size corresponds to a prioritized list of AWS instance types. For example, spot-micro may try to provision:

    • t3a.micro → t3.micro → t2.micro

  6. Complete the Server Build
    Spot servers take the same time to provision as regular servers. They are marked with a Spot icon in the server list.

Viewing Costs and Savings

Once the server is running:

  • View your actual instance type under the server's Size field

  • Check cost breakdown and savings from the server details pane

While no upfront pricing is available, the cost will never exceed that of the equivalent regular instance (e.g., t3.micro for spot-micro).

Best Practices

  • Avoid manually shutting down from the OS

  • Back up important data externally if the workload cannot tolerate interruption

  • Monitor cost and savings periodically in the server view

Need Help?

Email: spinup@yale.edu
Slack: #spinup channel on the Yale workspace