Doc 1144

cloud/file-shares/configure-file-shares.mdx

@@ -15,7 +15,7 @@
 </Info>
 
 <Tip>
-  The best practice is to create VAST shares **when** creating [GPU clusters](/edge-ai/ai-infrastructure/create-an-ai-cluster) or **before** provisioning the corresponding [compute resources](/cloud/virtual-instances/types-of-virtual-machines) (such as VMs).
+  The best practice is to create VAST shares **when** creating [GPU clusters](/edge-ai/ai-infrastructure/create-a-bare-metal-gpu-cluster) or **before** provisioning the corresponding [compute resources](/cloud/virtual-instances/types-of-virtual-machines) (such as VMs).
 </Tip>
 
 The creation flow for both types starts the same. Use the steps below and follow the instructions for the selected file share type.
@@ -42,7 +42,7 @@
   Manually change the OS settings' existing Bare Metal network interface.
 </Info>
 
 ## Configure file shares for Linux VMs and Bare Metal
 
 This section describes creating and connecting a standard NFS-based file share using a private network. It can be used with Linux virtual machines or bare-metal servers.
 
@@ -56,9 +56,9 @@
 
 2. In the **Basic settings** panel, enter _File Share name_, specify _Size_, and select **Standard** as the _File Share type_.
 
 ![File Shares Standard 4 Pn](/images/file-shares-standard-4.png)
 
 3. In the **Network settings** panel, select the private _Network_ and _Subnetwork_ to use for the file share.
 4. In the **Access** panel, click **Add rule** and specify the IP addresses of the resources that should have access to the file share, and their access modes.
 5. Set **Additional options**, if required.
 
@@ -107,7 +107,7 @@
 </Info>
 
 <Tip>
-  The best practice is to create VAST shares **when** creating [GPU clusters](/edge-ai/ai-infrastructure/create-an-ai-cluster) or **before** provisioning the corresponding [compute resources](/cloud/virtual-instances/types-of-virtual-machines) (such as VMs).
+  The best practice is to create VAST shares **when** creating [GPU clusters](/edge-ai/ai-infrastructure/create-a-bare-metal-gpu-cluster) or **before** provisioning the corresponding [compute resources](/cloud/virtual-instances/types-of-virtual-machines) (such as VMs).
 </Tip>
 
 ### Step 1. Create a VAST share
@@ -116,7 +116,7 @@
 2. In the **Basic settings** panel, enter _File Share name_, specify _Size_, and select **VAST** as the _File Share type_.
 3. Set **Additional options**, if required.
 
 ![File Share Vast Pn](/images/file-share-vast.png)
 
 When VAST share type is selected, controls in **Network settings** and **Access** panels are disabled, and the network is assigned automatically.
 
@@ -139,19 +139,19 @@
 
 The VAST network only becomes available after the file share has been created. It is a third, dedicated network, separate and distinct from public and private networks.
 
 ![File Share Details Pn](/images/file-share-details.png)
 
 #### Adding VAST interface to an existing compute resource
 
 While the VAST interface can be attached to an already-provisioned GPU cluster or compute resource, this requires additional manual network configuration and is not the standard workflow.
 
 <Tip>
-  The best practice is to create VAST shares **when** creating [GPU clusters](/edge-ai/ai-infrastructure/create-an-ai-cluster) or **before** provisioning the corresponding [compute resources](/cloud/virtual-instances/types-of-virtual-machines) (such as VMs).
+  The best practice is to create VAST shares **when** creating [GPU clusters](/edge-ai/ai-infrastructure/create-a-bare-metal-gpu-cluster) or **before** provisioning the corresponding [compute resources](/cloud/virtual-instances/types-of-virtual-machines) (such as VMs).
 </Tip>
 
 **Attach VAST network interface**
 
-1. Go to server **Resource** settings ([VM](/cloud/virtual-instances/create-an-instance), [Bare Metal](/cloud/bare-metal-servers/create-a-bare-metal-server), or [GPU cluster](/edge-ai/ai-infrastructure/create-an-ai-cluster)).
+1. Go to server **Resource** settings ([VM](/cloud/virtual-instances/create-an-instance), [Bare Metal](/cloud/bare-metal-servers/create-a-bare-metal-server), or [GPU cluster](/edge-ai/ai-infrastructure/create-a-bare-metal-gpu-cluster)).
 2. Select the **Networking** tab and click **Add interface**.
 3. Click the **Network** drop-down and select the **VAST network**, then click **Add**.
 4. Once the interface is added, **note the following details** for use in subsequent steps:
@@ -172,7 +172,7 @@
 3. Note the **interface name** (such as `enp8s0` or `enp4s0`) for use in the following steps.
 4. Configure the interface as described below for the relevant instance type.
 
 ##### **Configuring VAST interface for VMs and GPU clusters**
 
 Replace `198.51.100.25` with the IP address from Step 2. Replace `enp8s0` with the interface name from above, if different.
 
@@ -313,7 +313,7 @@
   Always use NFS version 3 (vers=3) when mounting VAST file shares. If the system does not support the `nconnect` option, install the [VAST Enhanced NFS Client](https://vastnfs.vastdata.com/docs/4.0/download.html).
 </Info>
 
 ![File Share Mount Pn](/images/file-share-mount.png)
 
 ## Resizing file shares
 

docs.json

@@ -642,8 +642,10 @@
               {
                 "group": "GPU cloud",
                 "pages": [
-                  "edge-ai/ai-infrastructure/about-our-ai-infrastructure",
-                  "edge-ai/ai-infrastructure/create-an-ai-cluster"
+                  "edge-ai/ai-infrastructure/about-gpu-cloud",
+                  "edge-ai/ai-infrastructure/create-a-bare-metal-gpu-cluster",
+                  "edge-ai/ai-infrastructure/spot-bare-metal-gpu",
+                  "edge-ai/ai-infrastructure/manage-a-bare-metal-gpu-cluster"
                 ]
               },
               {
@@ -1623,15 +1625,19 @@
     },
     {
       "source": "/docs/cloud/ai-Infrustructure/about-our-ai-infrastructure",
-      "destination": "/docs/edge-ai/ai-infrastructure/about-our-ai-infrastructure"
+      "destination": "/docs/edge-ai/ai-infrastructure/about-gpu-cloud"
     },
     {
       "source": "/docs/cloud/ai-Infrustructure/create-an-ai-cluster",
-      "destination": "/docs/edge-ai/ai-infrastructure/create-an-ai-cluster"
+      "destination": "/docs/edge-ai/ai-infrastructure/create-a-bare-metal-gpu-cluster"
     },
     {
       "source": "/docs/cloud/ai-Infrustructure/about-virtual-vpod",
-      "destination": "/docs/edge-ai/ai-infrastructure/about-our-ai-infrastructure"
+      "destination": "/docs/edge-ai/ai-infrastructure/about-gpu-cloud"
+    },
+    {
+      "source": "/docs/edge-ai/ai-infrastructure/about-our-ai-infrastructure",
+      "destination": "/docs/edge-ai/ai-infrastructure/about-gpu-cloud"
     },
     {
       "source": "/docs/edge-ai/inference-at-the-edge/:slug*",

edge-ai/ai-infrastructure/about-gpu-cloud.mdx

@@ -0,0 +1,92 @@
+---
+title: About GPU Cloud
+sidebarTitle: About GPU Cloud
+---
+
+GPU Cloud provides dedicated compute infrastructure for machine learning workloads. Use GPU clusters to train models, run inference, and process large-scale AI tasks.
+
+## What is a GPU cluster
+
+A GPU cluster is a group of interconnected servers, each equipped with multiple high-performance GPUs. Clusters are designed for workloads that require massive parallel processing power, such as training large language models (LLMs), fine-tuning foundation models, running inference at scale, and high-performance computing (HPC) tasks.
+
+<Frame>
+  <img src="/images/docs/edge-ai/ai-infrastructure/about-gpu-cloud/create-cluster-page.png" alt="GPU Cloud create cluster page showing region selection, cluster type, and GPU configuration options" />
+</Frame>
+
+All nodes in a cluster share the same configuration: operating system image, network settings, and storage mounts. This ensures consistent behavior across the cluster.
+
+## Cluster types
+
+Gcore offers two types of GPU clusters:
+
+| Type | Description | Best for |
+|------|-------------|----------|
+| **Bare Metal GPU** | Dedicated physical servers with guaranteed resources. No virtualization overhead | Production workloads, long-running training jobs, and latency-sensitive inference |
+| **Spot Bare Metal GPU** | Same hardware as Bare Metal, but at a reduced price (up to 50% discount). Instances can be preempted with a 24-hour notice when capacity is needed | Fault-tolerant training with checkpointing, batch processing, development, and testing |
+
+<Info>
+Spot instances are ideal for workloads that can handle interruptions. When a Spot cluster is reclaimed, you receive an email notification 24 hours before deletion. Use this time to save critical data to file shares or object storage.
+</Info>
+
+Clusters can scale to hundreds of nodes. Production deployments with 250+ nodes in a single cluster are supported, limited only by regional stock availability.
+
+## Available configurations
+
+Select a configuration based on your workload requirements:
+
+| Configuration | GPUs | Interconnect | RAM | Storage | Use case |
+|--------------|------|--------------|-----|---------|----------|
+| H100 with InfiniBand | 8x NVIDIA H100 80GB | 3.2 Tbit/s InfiniBand | 2TB | 8x 3.84TB NVMe | Distributed LLM training requiring high-speed inter-node communication |
+| H100 (bm3-ai-ndp) | 8x NVIDIA H100 80GB | 3.2 Tbit/s InfiniBand | 2TB | 6x 3.84TB NVMe | Distributed training and latency-sensitive inference at scale |
+| A100 with InfiniBand | 8x NVIDIA A100 80GB | 800 Gbit/s InfiniBand | 2TB | 8x 3.84TB NVMe | Multi-node ML training and HPC workloads |
+| A100 without InfiniBand | 8x NVIDIA A100 80GB | 2x 100 Gbit/s Ethernet | 2TB | 8x 3.84TB NVMe | Single-node training, inference for large models requiring more than 48GB VRAM |
+| L40S | 8x NVIDIA L40S | 2x 25 Gbit/s Ethernet | 2TB | 4x 7.68TB NVMe | Inference, fine-tuning small to medium models requiring less than 48GB VRAM |
+
+Outbound data transfer (egress) from GPU clusters is free. For pricing details, see [GPU Cloud billing](/edge-ai/billing).
+
+## InfiniBand networking
+
+InfiniBand is a high-bandwidth, low-latency interconnect technology used for communication between nodes in a cluster.
+
+InfiniBand is configured automatically when you create a cluster. If the selected configuration includes InfiniBand network cards, all nodes are placed in the same InfiniBand domain with no manual setup required.
+
+H100 configurations typically have 8 InfiniBand ports per node, each creating a dedicated network interface.
+
+InfiniBand matters most for distributed training, where models that don't fit on a single node require frequent gradient synchronization between GPUs. The same applies to multi-node inference when large models are split across servers. In these cases, InfiniBand reduces communication overhead significantly compared to Ethernet.
+
+For single-node workloads or independent batch jobs that don't require node-to-node communication, InfiniBand provides no benefit. Standard Ethernet configurations work equally well and may be more cost-effective.
+
+## Storage options
+
+GPU clusters support two storage types:
+
+| Storage type | Persistence | Performance | Use case |
+|-------------|-------------|-------------|----------|
+| Local NVMe | Temporary (deleted with cluster) | Highest IOPS, lowest latency | Training data cache, checkpoints during training |
+| File shares | Persistent (independent of cluster) | Network-attached, lower latency than object storage | Datasets, model weights, shared checkpoints |
+
+Learn more about [configuring file shares](/cloud/file-shares/configure-file-shares) for persistent storage and sharing data between nodes.
+
+## Cluster lifecycle
+
+```
+Create --> Configure --> Run workloads --> Resize (optional) --> Delete
+```
+
+1. **Create**: Select region, GPU type, number of nodes, image, and network settings when [creating a Bare Metal GPU cluster](/edge-ai/ai-infrastructure/create-a-bare-metal-gpu-cluster).
+
+2. **Configure**: Connect via SSH to each node, install required dependencies, and mount file shares to prepare the environment for workloads.
+
+3. **Run workloads**: Execute training jobs, run inference services, process data.
+
+4. **Resize**: Add or remove nodes based on demand. New nodes inherit the cluster configuration, which you can manage in the [Bare Metal GPU cluster details](/edge-ai/ai-infrastructure/manage-a-bare-metal-gpu-cluster).
+
+5. **Delete**: Remove the cluster when no longer needed. Local storage is erased; file shares remain.
+
+
+<Info>
+GPU clusters may take 15–40 minutes to provision, and their configuration (image, network, and storage) is fixed at creation. Local NVMe storage is temporary, so critical data should be saved to persistent file shares. Spot clusters can be interrupted with a 24-hour notice, and cluster size is limited by available regional stock.
+</Info>
+
+Hardware firewall support is available on servers equipped with BlueField network cards, enhancing network security for GPU clusters.
+