<img alt="" src="https://secure.insightful-enterprise-intelligence.com/783141.png" style="display:none;">

Access NVIDIA H100s from just $2.06/hour. Reserve here

Deploy 8 to 16,384 NVIDIA H100 SXM GPUs on the AI Supercloud. Learn More

|

Published on 12 Sep 2024

5 Tips to Maximise Your Hyperstack Experience

TABLE OF CONTENTS

updated

Updated: 12 Sep 2024

NVIDIA H100 GPUs On-Demand

Sign up/Login

Hyperstack is your go-to platform for flexible and budget-friendly solutions to deploy high-performance workloads in the cloud. From AI and machine learning to experimenting with the latest LLMs like Llama 3.1 70-B, Hyperstack provides a powerful environment designed to meet the demands of various advanced computing tasks.

No matter, if you are just getting started with Hyperstack or our regular user looking to maximise your productivity on Hyperstack, this blog is made for you. Continue reading to explore the top 5 ways to enhance your experience with Hyperstack. We are dedicated to providing you with maximum value and resources to help you make the most of our platform.

1. Ensure Connectivity to Virtual Machines

A successful cloud computing experience begins with ensuring that your virtual machines (VMs) are connected properly. Hyperstack uses the Secure Shell (SSH) protocol to provide a secure and encrypted connection between your local machine and the remote server. This connection allows you full command-line interface access to manage your VM effectively. However, certain issues might arise, causing interruptions or failed connections so it's important to follow the below steps:

Check VM Status

Before anything else, make sure your VM is active. Head to the "My Virtual Machines" page on Hyperstack and verify that the status reads “ACTIVE" status. Sometimes, it can take a few minutes for all services to initialise after deployment. If you encounter any connection errors, try again after a brief wait.

deploy-vm-10-7acd90d826caa05c7bbc2657c487a8c6

Assign a Public IP

One common mistake is forgetting to assign a public IP to your VM, which prevents SSH access. A public IP is required for a direct, remote connection to the internet, as private IPs are only accessible within local networks. To assign a public IP, go to the "Networking" tab on the VM’s details page and toggle "Public IP Attached."

Enable SSH Traffic

deploy-vm-11-ecd8c53003182041d3a2881d0010f6c6-1

Finally, you need to ensure that SSH traffic is properly enabled. To enable SSH access to your Hyperstack VMs, you must create a firewall rule that permits incoming traffic on port 22 (the default SSH port) from any IP address (0.0.0.0/0). Check out the steps below:

  1. Go to the 'Virtual Machines' section in Hyperstack and select the VM you want to manage.
  2. Once the VM is marked ACTIVE, go to the 'Firewall' tab in its details.
  3. Click the 'Enable SSH access' button.

This will automatically create a firewall rule and allow incoming traffic from any IP address (0.0.0.0/0) on port 22 via TCP for secure SSH access.

For Windows VMs, you can connect using RDP (Remote Desktop Protocol). Check out our documentation for detailed instructions.

2. Manage Your Resources Efficiently

One of the most effective ways to manage your resources on Hyperstack is by understanding the different states of your virtual machines and how they impact your billing. When you delete, stop or hibernate your VM, each action affects your resources and costs. To manage your resources efficiently, follow the below steps:

Deleting vs. Stopping vs. Hibernating Your VM

When your VM is not in use you have the option of deleting, stopping, or hibernating it.

  1. Delete: When you’re done using a VM, deleting it removes all associated costs and wipes the configuration data. If you want to pause your VM temporarily but retain the configuration, you have two options: stopping or hibernating the VM.
  2. Stop: Stopping your VM transitions it from an ACTIVE to a SHUT OFF state. However, the resources allocated remain reserved for your VM, meaning you continue to incur full billing costs for the resources you're not actively using.
  3. Hibernate: Hibernating your virtual machine allows you to save its current state to disk for quick restoration later. During this process, the VM’s configuration and root disk data are preserved to ensure a fast restart without needing a full boot-up. Meanwhile, resources like CPUs, GPUs, memory and ephemeral storage are deallocated, pausing billing until the VM is restored. If you're pausing a project or need to retain a VM without actively using it, hibernation offers the perfect solution. It's a great way to manage long-term costs and avoid paying for idle resources. Hibernations ensure you only pay for what you need, reducing overall costs for using Hyperstack virtual machines. Please note that charges will only accrue for the VM’s root disk data saved to persistent storage and the public IP address if attached.

For a comprehensive guide on managing your VM's state on Hyperstack, click here.

3. Choose the Right GPU and Flavor for Your Workload

Selecting the correct GPU for your AI workloads ensures your project runs smoothly. Hyperstack offers a variety of GPU-based instances, commonly referred to as "flavors." These flavors come with predefined combinations of GPU, CPU, RAM and storage. Our pre-configured flavors are designed to support a range of tasks from small-model training to large-scale AI workloads. To choose the right GPU and flavor, you need to:

Understand the Flavors

If you're working on a small project or training a simple model, you don’t need the most powerful GPU available. You can opt for a lighter flavor can save you resources and reduce costs. Similarly, if your workload involves large-scale LLM training or inference tasks, choosing a flavor with more powerful GPUs, such as the NVIDIA A100 or NVIDIA H100 SXM will better suit your needs. Hyperstack provides a wide range of GPU flavors designed to accommodate various tasks, so it's important to carefully match the flavor to your project's requirements.

Have You Tried Our GPU Selector?

Our GPU Selector for LLMs makes it easy to find the perfect GPU for your needs. With just a few clicks, you can get tailored recommendations based on your model, training approach and inference requirements.

4. Understand Data Storage and Backup

Data storage and backup are often overlooked but are critical to your AI project's success on Hyperstack. Understanding the differences between ephemeral storage and persistent storage can help you avoid data loss and ensure your critical files are always backed up.

Root Disk vs. Ephemeral Storage

The root disk on a Hyperstack VM contains the operating system and necessary system files. While the root disk persists between reboots, ephemeral storage is different. Ephemeral storage is used for active workloads and once a VM is terminated, all data stored on ephemeral storage is lost. This makes ephemeral storage ideal for temporary files, but it's essential to ensure that important data is stored elsewhere.

Use Persistent Storage Solutions

For data you cannot afford to lose, such as project files, databases or configuration data, always back it up to persistent storage. Hyperstack offers Shared Storage Volumes (SSVs) which provide a reliable, persistent solution to store your data. SSVs ensure that even if your VM is terminated, the data remains safe and accessible.

Note: As SSVs are network-based, they may be slower compared to root disks or ephemeral storage. 

5. Configure Firewalls for Secure Internet Access

Security should always be a top priority when working in the cloud. Hyperstack’s new virtual machines are secure by default, meaning that no incoming traffic is allowed unless you specifically configure the firewall to permit it. 

Default Firewall Configuration

By default, Hyperstack VMs come with pre-configured rules permitting outgoing IPv4 and IPv6 traffic using any internet protocol, on any port (1-65536), and to any destination IP address (0.0.0.0/0). If required, these pre-configured firewall rules for incoming traffic can be deleted to restrict outgoing traffic. However, no firewall rules for incoming traffic are enabled, which ensures that your VMs are secure by default.

Customise Firewall Rules

If you need to enable incoming traffic to your VM, you must create custom firewall rules according to your specific requirements. To create a firewall containing firewall rules that can be attached to any virtual machine, follow the steps below.

  1. In Hyperstack, click on the Firewalls section located on the sidebar and click Create a new firewall.

  2. Give your firewall a name and choose an environment to create it. OR provide a description.

  3. Add firewall rules to your firewall by clicking Add New Inbound/Outbound rule depending on your needs.

  4. Complete the fields for each firewall rule, including:

  1. Ether type: Select either IPv4 or IPv6.
  2. Internet protocol: Choose TCP, UDP, or ICMP.
  3. Port range: Specify the range of ports to allow (e.g., 1-65536).
  4. Remote IP prefix: Enter the IP address for permitted traffic.
  5. Click Add.

5. Repeat steps 3 and 4 for all the firewall rules you need.

You are now ready to attach this firewall to a virtual machine to apply its security configuration. To learn how, click here.

We hope these tips help you get the most out of Hyperstack and enhance your experience with us. Applying these practices to your projects could streamline your workflows, save on costs and fully leverage our cloud platform's powerful features.

Stay tuned for more such articles. Until next time.

Happy computing!

FAQs

How do I check if my VM is active on Hyperstack?

Go to the "My Virtual Machines" page on Hyperstack and ensure the status reads “ACTIVE.”

Why is my SSH connection failing?

Please ensure your VM has a public IP assigned and SSH traffic is enabled in the firewall setting to avoid a failed SSH connection.

What is the difference between stopping and hibernating a VM?

Stopping a VM incurs ongoing costs for reserved resources while hibernation saves state and pauses billing.

How can I choose the right GPU for my workload?

You must match the GPU flavor to your project’s requirements, with lighter flavors for small tasks and powerful ones like NVIDIA H100 SXM for larger workloads. You can also check out our guide to choosing the right GPU. 

Subscribe to Hyperstack!

Enter your email to get updates to your inbox every week

Get Started

Ready to build the next big thing in AI?

Sign up now
Talk to an expert

Share On Social Media

Hyperstack - Tutorials link

3 Oct 2024

Meta has finally released its latest Llama 3.2 model designed to push the boundaries of ...

Hyperstack - Tutorials link

18 Sep 2024

We’re excited to announce the release of on-demand high-speed networking with SR-IOV. ...