Skip to main content

Quick Start

Prerequisites​

Before launching an instance, ensure you have:

  • Active Account: Access to the Tir Cloud Portal.
  • Sufficient Credits: On-Demand instances require minimum 1 hour credits; Committed instances require full commitment period credits upfront.

How to Create an Instance​

Step 1: Navigate to Instances​

  1. Log in to the Tir Cloud Portal.
  2. Select your Project from the left hand sidebar.
  3. Click on Instances under the Products menu and select Instance.

Step 2: Choose an Image​

Select an image that matches your framework and workload requirements. TIR provides three types of instance images: Pre-built, Base OS, and Custom.

To understand the differences and usage of each image type, see the Instance Image documentation. For the complete list of supported images available on the platform, refer to the Supported Image List.

JupyterLab Support​

Enable "JupyterLab Supported" only if your selected image includes JupyterLab. This option is applicable for images built using the TIR Image Builder or TIR prebuilt images with JupyterLab support. Always verify compatibility before enabling.

Step 3: Configure Compute Resources​

Select your instance type (GPU, CPU, Spot, or Private Cluster), choose the hardware plan (e.g., NVIDIA H200 SXM), and then pick the appropriate vCPU/RAM configuration (e.g., 1 GPU / 30 vCPUs / 375GB RAM).

Step 4: Pricing & Storage​

Choose between On-Demand (hourly) or a Committed Plan (1, 3, or 6 months) based on your workload duration and cost preferences. Storage starts with 30GB free and can be expanded up to 15TB.

Step 5: Network & Security​

  1. SSH Key: Select an existing key or create a new one from the dropdown. If you create a new key, the private key will be downloaded as a .pem file — store it securely, as it cannot be re-downloaded. See SSH Keys Management for connection instructions.
  2. Security Group: Attach a predefined group to control network access.

Step 6: Advanced Settings (Optional)​

SettingDescription
Parallel / Shared File SystemsMount SFS or Dataset storage for large scale data access
Startup ScriptsAutomate environment setup on boot
Static IPReserve a persistent IP address for the instance
Storage EncryptionEnable AES-256 encryption for disk

Step 7: Launch​

Review your configuration and click Launch Instance.

You can also create an instance via an API request. Refer to the Instance Creation API documentation for request parameters and examples.

Key Options to Know​

  • SSH Access: Enable SSH by attaching an SSH public key during instance creation. A Security Group with port 22 open must be attached for SSH to function. If you missed adding a key during creation, you can enable SSH Access afterward and upload the key, provided a Security Group is already attached.

  • Workspace Path: The instance workspace is mounted at /home/jovyan. Always save persistent files within this path. Files stored outside this directory may be lost on container restart.

  • Disk Size: Default storage is 30GB (free), expandable up to 5TB. For requirements above 5TB, raise a support ticket. You can also enable Storage Encryption for AES-256 disk-level encryption.

Launch App​

The Launch App button opens an interactive workspace for managing files, writing and executing code, and creating documentation.

Example: To verify GPU resources on an H100 instance, run nvidia-smi inside a Python 3 kernel.

Instances Event​

The Instances Event log provides a chronological audit trail of the instances's lifecycle. It tracks system messages, reasons for state changes (e.g., starting, stopping, or errors), and the age of the event to help with troubleshooting and monitoring.

Precautions​

ActionWhat to Know
Container restartFiles outside /home/jovyan may be lost
Deleting an instancePermanently deletes the associated disk and workspace. Back up data before deletion
SSH key changesRequires a Security Group to be attached before uploading a new key

Best Practices​

  • Start with a prebuilt Jupyter image for faster onboarding.
  • Store large datasets in EOS buckets and mount them rather than keeping them on instance disk.
  • Use Committed Plans for long running or regular workloads.
  • Always back up your SSH key as it is the recommended way to access the instance shell.