Resource Management

Configure CPU, memory, disk, and GPU resources for Daytona sandboxes

Overview

Daytona allows you to allocate specific compute resources to each sandbox. Resources are defined when creating a sandbox and determine its computational capacity.

Resource Types

Available Resources

Resource	Type	Unit	Description
cpu	number	Cores	Number of CPU cores allocated
memory	number	GiB	RAM allocation in gibibytes
disk	number	GiB	Disk space allocation in gibibytes
gpu	number	Units	Number of GPU units allocated

Setting Resources

Basic Configuration

Resources are specified in the resources parameter when creating a sandbox:

TypeScript

import { Daytona } from '@daytonaio/sdk'

const daytona = new Daytona()

const sandbox = await daytona.create({
  image: 'python:3.12',
  resources: {
    cpu: 4,      // 4 CPU cores
    memory: 8,   // 8 GiB RAM
    disk: 50     // 50 GiB disk space
  }
})

console.log(`CPU: ${sandbox.cpu} cores`)
console.log(`Memory: ${sandbox.memory} GiB`)
console.log(`Disk: ${sandbox.disk} GiB`)

GPU Allocation

For GPU-accelerated workloads:

TypeScript

const sandbox = await daytona.create({
  image: 'tensorflow/tensorflow:latest-gpu',
  resources: {
    cpu: 8,
    memory: 32,
    disk: 100,
    gpu: 1  // 1 GPU unit
  }
})

console.log(`GPU: ${sandbox.gpu} units`)

Default Resources

When resources are not specified:

• Sandboxes use default resource allocations based on your organization’s configuration
• Resources can vary depending on the region and availability

TypeScript

// Uses default resources
const sandbox = await daytona.create({
  snapshot: 'default-python'
})

console.log(`Default CPU: ${sandbox.cpu} cores`)
console.log(`Default Memory: ${sandbox.memory} GiB`)

Resource Planning

Use Case Examples

Lightweight Development

const devSandbox = await daytona.create({
  resources: {
    cpu: 1,
    memory: 2,
    disk: 10
  }
})

Standard Application Testing

const testSandbox = await daytona.create({
  resources: {
    cpu: 2,
    memory: 4,
    disk: 20
  }
})

Data Science Workloads

const dataScienceSandbox = await daytona.create({
  image: Image.debianSlim('3.12')
    .pipInstall(['pandas', 'numpy', 'scikit-learn']),
  resources: {
    cpu: 8,
    memory: 32,
    disk: 100
  }
})

Machine Learning Training

const mlSandbox = await daytona.create({
  image: 'tensorflow/tensorflow:latest-gpu',
  resources: {
    cpu: 16,
    memory: 64,
    disk: 200,
    gpu: 2
  }
})

Resizing Sandboxes

Sandbox resizing allows you to adjust resources without recreating the sandbox.

TypeScript

// Resize an existing sandbox
await sandbox.resize({
  cpu: 8,
  memory: 16,
  disk: 100
})

await sandbox.refreshData()
console.log(`New CPU: ${sandbox.cpu} cores`)
console.log(`New Memory: ${sandbox.memory} GiB`)

Monitoring Resources

Check Current Allocation

TypeScript

const sandbox = await daytona.get('sandbox-id')

console.log('Resource Allocation:')
console.log(`  CPU: ${sandbox.cpu} cores`)
console.log(`  Memory: ${sandbox.memory} GiB`)
console.log(`  Disk: ${sandbox.disk} GiB`)
console.log(`  GPU: ${sandbox.gpu || 0} units`)

List Sandboxes with Resources

TypeScript

const result = await daytona.list()

for (const sandbox of result.items) {
  console.log(`${sandbox.id}:`)
  console.log(`  CPU: ${sandbox.cpu}, Memory: ${sandbox.memory}, Disk: ${sandbox.disk}`)
}

Snapshots with Resources

When creating snapshots with specific resource requirements:

TypeScript

import { Daytona, Image } from '@daytonaio/sdk'

const daytona = new Daytona()

const image = Image.debianSlim('3.12')
  .pipInstall(['pandas', 'numpy', 'matplotlib'])

await daytona.snapshot.create(
  {
    name: 'data-science-env',
    image,
    resources: {
      cpu: 4,
      memory: 16,
      disk: 50
    }
  },
  {
    onLogs: console.log
  }
)

// Create sandboxes from snapshot (inherits resource allocation)
const sandbox = await daytona.create({
  snapshot: 'data-science-env'
})

Best Practices

1. Start small: Begin with minimal resources and scale up based on actual usage.

2. Match workload to resources:

   • CPU-intensive: Increase CPU cores
   • Memory-intensive: Increase RAM
   • Data processing: Increase disk space
   • ML/AI workloads: Add GPU resources

3. Use cost-effective configurations: Right-size resources to avoid over-provisioning.

4. Monitor and adjust: Use sandbox resizing to optimize resources over time.

5. Snapshot resource templates: Create snapshots with different resource profiles for common use cases.

Resource Limits

Resource availability may be subject to:

• Organization quotas
• Region availability
• Plan limitations

Contact your organization administrator or Daytona support for quota increases.

Related

• Custom Images - Configure custom Docker images
• Auto Lifecycle - Automatic resource cleanup
• Regions - Region-specific resource availability