10 Compute ‐ basic - KleberVales/OCI-Architect-Associate GitHub Wiki

1. Compute Service

1.1 Introduction

1.2 Understanding Shapes

1.3 Types of Compute Instances

1.4 Important: Data Persistence

2. Platform and Custom Images

2.1 Introduction

2.2 What is an Instance Image?

2.3 Image Sources

2.4 Platform Images

2.5 Custom Images: Key Details

2.6 Windows Custom Images

3. Import Export and BYOI

3.1 Introduction

3.2 Image Import/Export

3.3 Launch Modes for Importing Images

3.4 Steps to Replicate an Image Across Regions

3.5 Bring Your Own Image (BYOI)

4. Compute Shapes

4.1 Introduction

4.2 What is a Shape?

4.3 Two Options for Shapes

4.4 Understanding Flexible Shapes

4.5 Benefits of Flexible Shapes

5. GPU Shapes

5.1 Introduction

5.2 What are GPU Shapes?

5.3 NVIDIA GPU Support

5.4 Use Cases for GPU Shapes

5.5 Cluster Networking

5.6 GPU Instance Options

6. Bare Metal vs Virtual Machine vs Dedicated Hosts

6.1 Introduction

6.2 Bare Metal Instances

6.3 Virtual Machine Instances

6.4 Dedicated Hosts

6.5 Comparison at a Glance

7. Capacity Types Preemptible Instances

7.1 Introduction

7.2 What are Preemptible Instances?

7.3 Restrictions and Limitations

7.4 Termination Notices

7.5 Use Cases

7.6 Creating a Preemptible Instance

8. Capacity Types Capacity Reservations

8.1 Introduction

8.2 What Is a Capacity Reservation?

8.3 Creating a Capacity Reservation

8.4 Example Scenarios Where Capacity Reservations Help

8.5 Support and Limitations

9. Capacity Types Dedicated VM Host

9.1 Introduction

9.2 What is a Dedicated Virtual Machine Host?

9.3 Billing Model

9.4 Choosing a Shape

9.5 Use Cases

9.6 Dedicated Virtual Machine Host Shapes

9.7 Limitations

10. Dedicated VM Host

10.1 Introduction

10.2 Why Use Dedicated Virtual Machine Hosts?

10.3 Creating a Dedicated Virtual Machine Host

10.4 Exploring the Dedicated VM Host

10.5 Creating an Instance on the Host

10.6 Verifying the Instance and Host

10.7 Editing and Resizing the Instance

10.8 Moving an Instance Off the Dedicated Host

.
.
.
.
.

1. Compute Service

The OCI Compute Service is a foundational offering that enables you to provision and manage compute hosts called instances. It provides a range of shapes—templates defining OCPUs, memory, and other resources—to suit different compute and application needs. OCI offers both bare metal instances, which give you dedicated physical server access, and virtual machine instances, which run as independent environments on shared physical hardware. A key operational detail to remember is that any changes made to an instance's local drives are lost upon termination, while changes saved to attached volumes are retained. Understanding these fundamentals is essential for effectively using OCI Compute Service in your cloud environment.

1.1 Introduction

Welcome to this lesson on the OCI Compute Service. My name is Samvit Mishra, and I'm a senior principal training lead. In this lesson, we'll explore what the Compute Service is, how it works, and the key concepts you need to understand.

1.2 Understanding Shapes

OCI offers a variety of shapes designed to meet a wide range of compute and application requirements. But what exactly is a shape?

A shape is a template that defines:

The number of OCPUs (Oracle CPUs)

The amount of memory

Other resources allocated to an instance

In other words, when you choose a shape, you're deciding how much compute power and memory your instance will have.

1.3 Types of Compute Instances

OCI Compute Service offers two main types of instances:

1. Bare Metal Instances

Provide dedicated physical server access
You get the entire physical machine to yourself

2. Virtual Machine (VM) Instances

An independent computing environment that runs on top of physical bare metal hardware
Multiple VMs can share the same underlying hardware

1.4 Important: Data Persistence

There's a critical distinction to remember when working with instances:

Storage Type	Behavior on Instance Termination
Local drives	Changes are lost when the instance is terminated
Attached volumes	Changes are retained when the instance is terminated

This means you should always save important data to attached volumes if you want it to persist beyond the life of the instance.

2. Platform and Custom Images

In this lesson, we covered the essentials of platform and custom images in OCI. When creating a compute instance, you must specify both a shape and an image—the image acts as a template for the virtual hard drive, determining the operating system and software. OCI offers multiple image sources: platform images (pre-built operating systems like Oracle Linux, Ubuntu, CentOS, Windows, and Oracle Autonomous Linux), Oracle images (enterprise solutions), custom images (created from existing OCI instances), partner images (third-party images from the Marketplace), and bring-your-own-image (importing your own OS versions, with a 400 GB maximum size limit). When creating a custom image from a running instance, the instance shuts down temporarily and restarts once complete, and any instances launched from that image will inherit the captured customizations. For Windows custom images, you can choose between generalized images (cleaned of instance-specific data) and specialized images (point-in-time snapshots useful for backups). Understanding these image types and their behaviors is essential for efficiently deploying and managing compute instances in OCI.

2.1 Introduction

Welcome to this lesson on platform and custom images. In this lesson, we'll explore how images work in Oracle Cloud Infrastructure, the different sources of images available to you, and some important considerations when working with custom images. Let's get started.

2.2 What is an Instance Image?

When you create a compute instance, you need to specify two things:

Shape – defines the compute resources (OCPUs, memory, etc.)
Image – determines the operating system and other software for an instance

You can think of an image as a template of a virtual hard drive. It provides the foundational software environment that your instance will run on.

2.3 Image Sources

OCI provides several sources for images. Let's look at each one:

1. Platform Images

Pre-built operating systems for Oracle Cloud Infrastructure
Provided directly by Oracle

2. Oracle Images

Oracle Enterprise images and solutions enabled for OCI

3. Custom Images

Created from existing instances running in OCI
Capture your own configurations and software

4. Partner Images

Trusted third-party images published in the Marketplace by Oracle partners

5. Bring Your Own Image (BYOI)

Using the custom image feature, you can import custom images into the OCI environment
Allows you to bring your own versions of operating systems to the cloud

2.4 Platform Images: Examples

Oracle provides several pre-built operating systems for OCI, including:

Oracle Linux
Ubuntu
CentOS
Microsoft Windows
Oracle Autonomous Linux

For a detailed list, visit the OCI documentation.

2.5 Custom Images: Key Details

Creating a Custom Image from a Running Instance

The instance shuts down and remains unavailable for several minutes
The instance restarts when the process completes
Instances launched from your custom image will include the customization, configuration, and software installed when you created the image

Bring Your Own Image (BYOI)

Enables you to bring your own versions of operating systems to the cloud
Requirement: The underlying hardware must support it
Limitation: The maximum image size that can be imported is 400 GB

2.6 Windows Custom Images

When creating a Windows custom image, there is support for two kinds of images:

Image Type	Description	Use Case
Generalized	Images that have been cleaned of instance-specific information	Standard deployments
Specialized	Point-in-time snapshots of the boot disk of a running instance	Creating backups of an instance

3. Import Export and BYOI

In this lesson, we explored image import/export and the Bring Your Own Image feature in OCI. Image import/export allows you to share custom images across tenancies and regions, using OCI Object Storage as the mechanism for replication. When importing images, you can choose from three launch modes: emulation mode (for VMs without paravirtualized drivers created outside OCI), paravirtualized mode (for VMs with paravirtualized drivers created outside OCI), and native mode (for images exported from OCI itself). To replicate an image across regions, you export it to a bucket in the source region, copy it to a bucket in the destination region, and then import it using the object's URL path. The Bring Your Own Image feature lets you export VMs from your existing virtualization environment and import them directly into OCI as custom images in either qcow2 or VMDK format, after which you can launch instances from them. BYOI supports cloud migration projects, works with both old and new operating systems, encourages experimentation, and increases infrastructure flexibility—though you must always comply with licensing requirements.

3.1 Introduction

Welcome to this lesson on Import/Export and Bring Your Own Image. In this lesson, we'll cover how to share custom images across tenancies and regions, the different launch modes available for importing images, the steps to replicate an image across regions, and how the Bring Your Own Image feature works. Let's get started.

3.2 Image Import/Export

You can share custom images across tenancies and regions using image import/export. This is a powerful capability that allows you to distribute your customized images wherever you need them.

Region Replication and Object Storage

When you replicate an image across a region, you leverage OCI Object Storage as the transport mechanism. The image is exported to object storage, copied to the destination region's bucket, and then imported there.

3.3 Launch Modes for Importing Images

Importing an image supports three different launch modes. Choosing the right mode depends on where the image came from and what drivers it supports.

Launch Mode	When to Use It
Emulation Mode	For virtual machines that don't support paravirtualized drivers and were created outside of OCI — such as older on-premises physical or virtual machines
Paravirtualized Mode	For virtual machines that support paravirtualized drivers but were created outside of the OCI environment
Native Mode	For images that were exported from Oracle Cloud Infrastructure

3.4 Steps to Replicate an Image Across Regions

If you want to replicate an image from one region to another, follow these three steps:

Export the image to an object storage bucket in the same region as the image
Copy the image to an object storage bucket in the destination region
Import the image in the destination region using the URL path to the image object

3.5 Bring Your Own Image (BYOI)

The Bring Your Own Image feature allows you to export virtual machines from your existing virtualization environment and import them directly into OCI as a custom image.

Supported Formats

You can import images in either of the following formats:

qcow2
VMDK

Once the image is imported, you can create an instance from that custom image.

Benefits of BYOI

The Bring Your Own Image feature provides several advantages:

Enables virtual machine cloud migration projects
Supports both old and new operating systems
Encourages experimentation
Increases infrastructure flexibility

Important Note

You must comply with all licensing requirements when using the Bring Your Own Image feature.

4. Compute Shapes

In this lesson, we covered compute shapes in OCI. A shape is a template that defines the number of OCPUs, amount of memory, and other resources allocated to an instance, and shapes are available with AMD, Intel, and ARM-based processors. There are two categories of shapes: fixed shapes, where you cannot customize OCPUs or memory and can create either bare metal or virtual machine instances; and flexible shapes, where you can customize OCPUs and memory but can only create virtual machine instances. Flexible shapes allow you to adjust resources at launch or when resizing, with network bandwidth and vNIC count scaling proportionately with OCPUs. For example, with the VM.Standard3.Flex shape, you can select 1 to 32 OCPUs with up to 64 GB of memory per CPU—so choosing 4 OCPUs allows up to 256 GB of memory. The key benefit of flexible shapes is the ability to build VMs that match your workload, optimizing performance while minimizing cost.

4.1 Introduction

Welcome to this lesson on compute shapes. In this lesson, we'll explore what shapes are, the different processor types available, the two main categories of shapes (fixed and flexible), and the benefits of using flexible shapes. Let's get started.

4.2 What is a Shape?

A shape is a template that determines:

The number of OCPUs
The amount of memory
Other resources allocated to an instance

Processor Types

Compute shapes are available with three processor types:

AMD processors
Intel processors
ARM-based processors

4.3 Two Options for Shapes

OCI offers two categories of shapes, each with different capabilities.

1. Fixed Shapes

You cannot customize the number of OCPUs or the amount of memory
You can create either a bare metal instance or a virtual machine instance using fixed shapes

2. Flexible Shapes

You can customize the number of OCPUs and the amount of memory
You can only create a virtual machine instance using flexible shapes
The maximum number of OCPUs and maximum amount of memory depend on the processor type

4.4 Understanding Flexible Shapes

Flexible shapes give you control over the resources allocated to your VM at launch or when resizing.

Key Characteristics

Lets you customize the number of OCPUs and amount of memory when launching or resizing your VM
Network bandwidth and the number of vNICs scale proportionately with the number of OCPUs

Naming Convention

Flexible shapes follow a naming pattern, for example:

VM.Standard.E3.Flex
VM.Standard3.Flex

Example: VM.Standard3.Flex

Let's walk through a practical example using the flexible shape VM.Standard3.Flex:

Parameter	Range
OCPUs	1 to 32 (customizable)
Maximum memory per CPU	64 GB

Example calculation:

If you select 4 OCPUs, the maximum memory you can select is 256 GB (4 × 64 GB)

This demonstrates how flexible shapes let you scale memory in proportion to the OCPUs you choose.

4.5 Benefits of Flexible Shapes

The primary benefit of flexible shapes is:

You can build VMs that match your workload, enabling you to optimize performance and minimize cost.

Instead of being locked into a fixed resource configuration, you can tailor your instance to precisely fit your application's needs.

5. GPU Shapes

In this lesson, we covered Compute GPU shapes in OCI. GPU shapes are designed for hardware-accelerated workloads—where applications offload certain tasks to specialized hardware to accelerate performance. OCI offers high-value, high-performance bare metal and virtual machine instances powered by NVIDIA GPUs, including models such as the A100, A10, V100, and P100. These shapes support use cases ranging from mainstream graphics and video to the most demanding AI training and high performance computing workloads. Oracle's ultra-low latency cluster networking, based on RDMA, delivers microsecond-level latency and allows customers to host more than 500 GPUs in a cluster without compromising performance. When choosing GPU instances, you can select from a variety of virtual machine options (featuring NVIDIA Ampere, Volta, and Pascal GPUs) or bare metal options (combining Intel or AMD CPUs with NVIDIA graphics processors, with some supporting cluster networking). Understanding GPU shapes is essential for workloads that require hardware acceleration and maximum compute performance.

5.1 Introduction

Welcome to this lesson on Compute GPU shapes. In this lesson, we'll explore what GPU shapes are, what hardware acceleration means, the GPU models OCI offers, common use cases, and how cluster networking enhances GPU performance. Let's get started.

5.2 What are GPU Shapes?

Oracle Cloud Infrastructure offers a variety of shapes designed to meet a range of compute and application requirements. One such shape is the GPU shape.

GPU shapes are designed for hardware-accelerated workloads.

What is Hardware Acceleration?

Hardware acceleration is a process where applications offload certain tasks to hardware in your system.

Purpose: To accelerate that task, achieving better performance than software alone could provide.

5.3 NVIDIA GPU Support

OCI offers the highest value and performance for bare metal and virtual machine compute instances that are powered by NVIDIA GPUs.

GPU Models Available

Some of the NVIDIA GPU models offered include:

A100
A10
V100
P100

5.4 Use Cases for GPU Shapes

GPU shapes are suited for a wide range of workloads, including:

Mainstream graphics and video
The most demanding AI training
High performance compute (HPC) workloads

5.5 Cluster Networking

Oracle's ultra-low latency cluster networking, based on RDMA, provides significant performance benefits:

Microsecond-level latency
Customers can host more than 500 GPUs in a cluster — without compromising on performance

5.6 GPU Instance Options

You can choose from a variety of virtual machine and bare metal compute instances.

Virtual Machine GPU Options

For VMs, you can choose from NVIDIA GPU families:

NVIDIA Ampere
NVIDIA Volta
NVIDIA Pascal

There are several VM GPU series available to select from.

Bare Metal GPU Options

GPU shapes include Intel or AMD CPUs combined with NVIDIA graphics processors
Some bare metal GPU shapes support cluster networking
Several bare metal GPU series are available to choose from

6. Bare Metal vs Virtual Machine vs Dedicated Hosts

In this lesson, we explored three compute options in OCI: bare metal instances, virtual machines, and dedicated hosts. Bare metal instances give you direct access to the underlying hardware and provide a dedicated physical server for the highest performance and strong isolation—making them ideal for non-virtualized workloads, performance-intensive workloads, Bring Your Own License scenarios, and workloads requiring a specific hypervisor. Virtual machine instances run on top of bare metal hardware, with a hypervisor virtualizing the server into smaller VMs; they are best suited for applications that don't require the performance and resources of an entire physical machine. Dedicated hosts combine both concepts—multiple VMs run on bare metal, but the entire server is dedicated to a single host—enabling you to meet compliance and regulatory requirements for isolation that would prevent you from using shared infrastructure. Understanding the differences between these options helps you choose the right compute environment for your workload's performance, licensing, and compliance needs.

6.1 Introduction

Welcome to this lesson on bare metal instances, virtual machines, and dedicated hosts. In this lesson, we'll explore each of these compute options, their key characteristics, and the scenarios where each is best suited. Let's get started.

6.2 Bare Metal Instances

Key Characteristics

You get direct access to the underlying hardware
Provides a dedicated physical server for highest performance and strong isolation

When to Use Bare Metal Instances

Bare metal instances are ideal for scenarios such as:

Workloads that are not virtualized
Performance-intensive workloads
Workloads that require Bring Your Own License (BYOL)
Workloads that require a specific hypervisor

6.3 Virtual Machine Instances

Key Characteristics

Virtual machine instances run on top of bare metal hardware
A hypervisor sits on top of the bare metal server to virtualize it into smaller VMs

When to Use Virtual Machines

Virtual machines are ideal for running applications that do not require the performance and resources of an entire physical machine.

6.4 Dedicated Hosts

Key Characteristics

A combination of bare metal and virtual machines
Multiple VMs run on bare metal, with the whole server dedicated to a single host

Why Use Dedicated Hosts?

This feature lets you meet compliance and regulatory requirements for isolation that would otherwise prevent you from using shared infrastructure.

6.5 Comparison at a Glance

Option	Description	Best For
Bare Metal	Dedicated physical server with direct hardware access	Non-virtualized, performance-intensive, BYOL, or specific hypervisor workloads
Virtual Machine	Runs on bare metal via a hypervisor; smaller, virtualized environments	Applications that don't need an entire physical machine
Dedicated Host	Multiple VMs on bare metal, with the whole server dedicated to one host	Compliance and regulatory isolation requirements

7. Capacity Types Preemptible Instances

In this lesson, we covered preemptible instances in OCI. Preemptible instances are identical to on-demand instances except that OCI can terminate them when compute capacity is needed elsewhere. They run on unused on-demand capacity, which allows OCI to offer them at a 50% discount, but this also means they can be terminated at any time—making them unsuitable for long-running workloads and best suited for short-term usage such as batch jobs like video encoding. There are several restrictions: you can only change the instance name after launch (not the shape), you cannot convert a preemptible instance to on-demand, preemptible capacity does not support bare metal instances, and preemptible instances cannot be used to create instance configurations or added to instance pools. OCI sends a termination notice through the Event Service, with the preemption action event limited to 30 seconds before termination begins. Creating a preemptible instance follows the same workflow as on-demand instances—you simply select Preemptible Capacity instead of On-Demand Capacity in the Placement section of the console. Understanding preemptible instances allows you to take advantage of significant cost savings for workloads that can tolerate interruption.

7.1 Introduction

Welcome to this lesson on preemptible instances. In this lesson, we'll explore what preemptible instances are, how they differ from on-demand instances, their limitations, how OCI handles termination, and some practical use cases. Let's get started.

7.2 What are Preemptible Instances?

Preemptible instances are the same as on-demand instances with one important exception:

If compute capacity is needed elsewhere, OCI can terminate those instances.

Key Characteristics

Run on unused on-demand capacity
OCI offers them at a 50% discount compared to on-demand instances
Can be terminated at any time
Not suitable for long-running workloads — designed for short-term usage

7.3 Restrictions and Limitations

There are several important restrictions to keep in mind when working with preemptible instances:

Instance Editing

When you edit an instance that uses preemptible capacity, only the name of the instance can be changed
You cannot change the shape of the instance after it is launched

Conversion

You cannot convert a preemptible instance to an on-demand instance

Pricing

Pricing is simple: it is always 50% cheaper than your on-demand rates

Capacity Reclamation

OCI can terminate preemptible instances if requests for on-demand instances increase, allowing the capacity to be reclaimed

Feature Limitations

Preemptible capacity does not support bare metal instances
You cannot use preemptible instances to create instance configurations
Preemptible instances cannot be used in instance pools

7.4 Termination Notices

OCI is transparent about when preemptible instances will be terminated:

OCI sends a termination notice through the Event Service before terminating a preemptible instance
The instance preemption action event is limited to 30 seconds before the instance termination begins

7.5 Use Cases

What are preemptible instances good for? Consider workloads that can tolerate interruption:

Batch jobs — for example, video encoding

These types of workloads can make excellent use of the cost savings preemptible instances offer.

7.6 Creating a Preemptible Instance

Preemptible instances use the same management experience as on-demand instances.

In the Console

To create a preemptible instance:

Navigate to the Placement section
Under Capacity Type, instead of selecting On-Demand Capacity, select Preemptible Capacity

That's the only change needed—everything else follows the standard instance creation workflow.

A demonstration of this process will be shown in the next lesson.

8. Capacity Types Capacity Reservations

8.1 Introduction

8.2 What are Preemptible Instances?

Preemptible instances are the same as on-demand instances with one important exception:

If compute capacity is needed elsewhere, OCI can terminate those instances.

Key Characteristics

Run on unused on-demand capacity
OCI offers them at a 50% discount compared to on-demand instances
Can be terminated at any time
Not suitable for long-running workloads — designed for short-term usage

8.3 Restrictions and Limitations

There are several important restrictions to keep in mind when working with preemptible instances:

Instance Editing

When you edit an instance that uses preemptible capacity, only the name of the instance can be changed
You cannot change the shape of the instance after it is launched

Conversion

You cannot convert a preemptible instance to an on-demand instance

Pricing

Pricing is simple: it is always 50% cheaper than your on-demand rates

Capacity Reclamation

OCI can terminate preemptible instances if requests for on-demand instances increase, allowing the capacity to be reclaimed

Feature Limitations

Preemptible capacity does not support bare metal instances
You cannot use preemptible instances to create instance configurations
Preemptible instances cannot be used in instance pools

8.4 Termination Notices

OCI is transparent about when preemptible instances will be terminated:

OCI sends a termination notice through the Event Service before terminating a preemptible instance
The instance preemption action event is limited to 30 seconds before the instance termination begins

8.5 Use Cases

What are preemptible instances good for? Consider workloads that can tolerate interruption:

Batch jobs — for example, video encoding

These types of workloads can make excellent use of the cost savings preemptible instances offer.

8.6 Creating a Preemptible Instance

Preemptible instances use the same management experience as on-demand instances.

In the Console

To create a preemptible instance:

Navigate to the Placement section
Under Capacity Type, instead of selecting On-Demand Capacity, select Preemptible Capacity

That's the only change needed—everything else follows the standard instance creation workflow.

A demonstration of this process will be shown in the next lesson.

9. Capacity Types Dedicated VM Host

In this lesson, we covered the dedicated virtual machine host capacity type in OCI. A dedicated virtual machine host lets you run compute VM instances on a dedicated, single-tenant server that is not shared with other customers. You are billed for the host itself as soon as you create it, but not for the individual VM instances placed on it. When creating a host, you select a shape that determines the available capacity and which instance types can be launched—keeping in mind that some OCPUs are reserved for VM management, so billed OCPUs differ from available OCPUs. Dedicated VM hosts are valuable for meeting compliance and regulatory isolation requirements and host-based licensing requirements that mandate licensing an entire server. Usable OCPUs and memory vary by shape—for example, DVH.Standard3.64 provides 60 usable OCPUs and 960 GB usable memory—and each host shape supports specific VM shapes, such as DVH.Standard2.52 supporting the VM.Standard2 series. Finally, while most compute features are supported, several are not, including autoscaling, capacity reservations, instance configurations, instance pools, burstable instances, and remote migration. Understanding these characteristics helps you determine when a dedicated VM host is the right choice for your workload's isolation and licensing needs.

9.1 Introduction

Welcome to this lesson on the dedicated virtual machine host capacity type. In this lesson, we'll explore what a dedicated virtual machine host is, how billing works, how to choose a shape, common use cases, available shapes, and the limitations to be aware of. Let's get started.

9.2 What is a Dedicated Virtual Machine Host?

A dedicated virtual machine host gives you the ability to run compute virtual machine instances on a dedicated server that is a single tenant and not shared with other customers.

Key Characteristics

Runs VM instances on a server dedicated to you alone
Provides single-tenant isolation — no sharing with other customers

9.3 Billing Model

Understanding the billing model is important when using dedicated virtual machine hosts:

You are billed for the dedicated virtual machine host as soon as you create it
You are not billed for the individual virtual machine instances that you place on it

In other words, you pay for the host itself, regardless of how many (or how few) VMs you run on it.

9.4 Choosing a Shape

When you create a dedicated virtual machine host, you select a shape for the host. This shape determines:

How much capacity is available
What type of instances can be launched on the host

Reserved OCPUs

Some OCPUs are reserved for virtual machine management.

Because of this, there will be a difference between the number listed for billed OCPUs compared to available OCPUs. The billed number reflects the total, while the available number is what remains after management overhead is reserved.

9.5 Use Cases

Dedicated virtual machine hosts are useful in scenarios such as:

Meeting compliance and regulatory isolation requirements — since you are not using shared infrastructure
Meeting host-based licensing requirements — where you must license an entire server

9.6 Dedicated Virtual Machine Host Shapes

The usable OCPUs and usable memory depend on the shape you select. Let's look at some examples:

Dedicated VM Host Shape	Usable OCPUs	Usable Memory
DVH.Standard3.64	60	960 GB

Supported Shapes for Hosted VMs

Each dedicated VM host shape supports specific VM shapes for the instances you place on it.

Example:

DVH.Standard2.52 supports the VM.Standard2 series

9.7 Limitations

Most compute features for virtual machine instances are supported for instances running on dedicated virtual machine hosts. However, the following features are not supported:

Autoscaling
Capacity reservations
Instance configurations
Instance pools
Burstable instances
Remote migration

Keep these limitations in mind when planning to use dedicated virtual machine hosts.

10. Dedicated VM Host

In this demonstration, we walked through the complete lifecycle of working with dedicated virtual machine hosts in OCI. We began by creating a dedicated VM host—naming it, selecting a compartment and availability domain, choosing a host shape (DVH.Standard3.64), and reviewing the compatible VM shapes it supports (VM.Standard3.Flex). After provisioning, we explored the host's details, monitoring, and hosted instances tabs. We then created an instance on the host, noting that placement is automatically tied to the dedicated host, that resources like OCPUs and memory are deducted from the host's total, and that boot and block volumes remain on the block volume infrastructure rather than the host itself. We verified resource accounting (960 GB total memory dropping to 928 GB, and 61 OCPUs dropping to 59 after allocating 32 GB and 2 OCPUs), then demonstrated editing the instance to add an OCPU, which required a reboot and further reduced the host's remaining OCPUs to 58. Finally, we showed how to move an instance off the dedicated host by terminating it (without deleting the boot volume) and recreating it using the same boot volume in the same availability domain, effectively migrating it to the regular public population. The key takeaways are that only compatible shapes can be launched on a dedicated host, and because disks live in the block volume infrastructure, instances can be moved on and off dedicated hosts as needed.

10.1 Introduction

Welcome to this demonstration on how to use the dedicated virtual machine host. In this lesson, we'll walk through the process of creating a dedicated virtual machine host, launching an instance on it, and then moving that instance off the host. Let's get started.

Note: This demonstration was created using an admin user. If you don't have an admin user, you will need to request the necessary permissions.

10.2 Why Use Dedicated Virtual Machine Hosts?

There will be times when you want your VMs to have isolation but don't want the hassle of managing a bare metal server. Dedicated virtual machine hosts let you:

Run VM instances on dedicated servers that are single tenant and not shared with other customers
Meet compliance and regulatory requirements for isolation

10.3 Creating a Dedicated Virtual Machine Host

Step 1: Navigate to the Service

Go to the navigation menu
Select Compute
Select Dedicated Virtual Machine Hosts
Click Create Virtual Machine Host to start the workflow

Step 2: Configure the Host

Setting	Value Used in Demo
Name	DVH-demo
Compartment	(Pre-selected — no change needed)
Availability Domain	Selected as required
Shape	DVH.Standard3.64

Step 3: Select the Shape

Click Edit shape to choose the host shape.

Important: If you get a server based on a fixed shape (like DVH.Standard2.52), you will only be able to provision shapes that are derivative from that family.

In this demo:

Host shape: DVH.Standard3.64
Compatible shapes: VM.Standard3.Flex

Step 4: Complete Placement and Create

Select a fault domain (e.g., FAULT-DOMAIN-1)
Add tags if needed (none in this demo)
Click Create to trigger provisioning

You can follow progress in Work requests until the host finishes provisioning.

10.4 Exploring the Dedicated VM Host

Once provisioning completes, click the Details tab to see:

Remaining OCPUs
Total amount of memory
Location
Host shape

Other Tabs

Monitoring — shows host performance (no data yet for a new host)
Host Instances — lists instances running on the host (empty initially)

10.5 Creating an Instance on the Host

Click Create instance to start the workflow.

Instance Configuration

Setting	Value
Name	from-dvh
Compartment	(Pre-selected)
Placement	Availability domain pre-selected to match the host; Dedicated virtual machine host pre-selected automatically
Shape	VM.Standard3.Flex (marked as compatible)

Adjusting Resources

When selecting the shape, you can choose the amount of OCPUs and memory you want. These resources are deducted from the host's total.

Networking and Storage

Security: No changes
VCN: demovcn01, public subnet
SSH keys: None (for demo purposes only)
Boot volumes: Still use block volume infrastructure, so size can be changed

Key point: No volumes—whether block or boot—are stored on the dedicated host.

Review and Create

Review the information and confirm:

Capacity type: Dedicated virtual machine host
Host: Shows which host will run the instance

The options are the same as an on-demand machine — the difference is it runs on your dedicated host, not shared with anyone.

Click Create.

10.6 Verifying the Instance and Host

Instance Details

Aside from a regular VM, the capacity type and host are what differ.

OCPU count: 3 (after an edit)
Network bandwidth: As allocated
RAM: 32 GB allocated
Block storage: Outside the VM — not a concern for this demo

Host Details

After creating the instance, check the host's Hosted instances tab:

The running instance appears here

Resource Accounting Example

Resource	Before	After (32 GB / 2 OCPU instance)
Total memory	960 GB	928 GB
Remaining OCPUs	61	59

10.7 Editing and Resizing the Instance

If you go to the instance's Actions → More Actions → Edit:

You cannot change the default domain because it's tied to the host machine's default domain
You can add more OCPUs

After saving, you'll be prompted to reboot. Once rebooted:

Instance now has 3 OCPUs
Host's remaining OCPUs drop from 59 to 58

Placement options are limited because the instance is tied to the host.

10.8 Moving an Instance Off the Dedicated Host

To move an instance off the dedicated host:

Terminate the instance — but do not check "Permanently delete the attached boot volume"
Click Create instance (e.g., named from-deleted)
Keep the same availability domain — because the block volume is tied to that availability domain
Under Change , select the boot volume from the previous instance (from the DVH)
Select the image — the shape will automatically change to something compatible
Complete the remaining settings (VCN, no SSH keys, boot volume as defined)
Click Create

Once created, the instance will be migrated from the dedicated host to the regular public population.

10.9 Key Takeaways from the Demo

You can only create instances based on the list of shapes allowed by the dedicated VM host's configuration and family
Disks are not held with the host — they remain with the block volume infrastructure
Because of this, you can delete an instance from the dedicated host and recreate it on the regular population if desired