Oracle Could Infrastructure Compute is a web service that provides bare metal and virtual machine (VM) compute capacity that delivers performance, flexibility, and control without compromise. It's powered by Oracle’s next generation internet-scale infrastructure service and is designed to help modern enterprises do more while paying less when developing and running their most demanding applications and workloads in the cloud.

Compute enables you to provision compute capacity in minutes through an easy-to-use web console. The bare metal compute instance, once provisioned, provides you access to the host. This gives you the flexibility, control, and performance without compromise needed for your most demanding applications and workloads, all while paying only for what you use.

Oracle Cloud Infrastructure is hosted in regions, each of which contain at least three availability domains. A region is simply a geographic area, such as “Germany” or “US West.” An availability domain is an isolated, fault-tolerant set of resources consisting of at least one data center. Availability domains don't share infrastructure such as a building, power, or cooling. A failure in one availability domain is unlikely to impact the availability of other availability domains.

A fault domain is a grouping of hardware and infrastructure within an availability domain. Fault domains let you distribute your instances so they're not on the same physical hardware within a single availability domain, thereby introducing another layer of fault tolerance. Each availability domain contains three fault domains. A hardware failure or maintenance on Compute hardware that affects one fault domain doesn't affect instances in other fault domains.

Oracle Cloud Infrastructure Compute offers the choice, flexibility, control, and performance that your applications and workloads need. You can provision compute instances in minutes through an easy-to-use web console or through an API. Oracle Cloud Infrastructure provides bare metal, virtual machine (VM), and GPU instances:

• Bare metal instances: A bare metal compute instance is a physical server that is fully dedicated to you. It helps address compliance requirements and saves money by leveraging server licenses that you already own, subject to your existing licensing terms. It provides uncompromising raw performance and includes instances with the latest generation NVMe SSDs. These instances offer over a million IOPS and are ideal for running any I/O-intensive application or big data workload. Bare metal compute instances run only the software that you choose, providing you with complete control. There is no Oracle applied software on your instance.
• VM instances: A VM compute instance is an isolated OS environment on a multitenant host. You can provision VM and bare metal instances in your virtual cloud network (VCN) side-by-side and manage them using the same console and API. VM compute instances offer a variety of shapes that let you tailor your deployment to a broad range of application and workload needs. This includes Dense I/O VMs that provide a high-performance instance type with large local non-volatile memory express (NVMe) SSD storage.
• GPU instances: GPU compute instances are available as either VMs or bare metal instances and can be deployed in your VCN side-by-side with others, and managed with the same console and API. They contain NVIDIA GPUs and are suitable for a GPU enabled and CUDA workloads. GPU compute instances offer a variety of shapes that let you tailor your deployment to your workload needs.
• HPC instances: HPC compute instances are available as bare metal instances and can be deployed in your VCN side-by-side with others, and managed with the same console and API. They contain high clock rate CPUs and are suitable for a variety of high-performance computing (HPC) workloads.

Compute instances, regardless of shape or size, launch within minutes from the time that you provision them from the Oracle Cloud Infrastructure Console or issue the LaunchInstanceRequest API command.

You can remotely connect to your instance by using the industry standard secure shell (SSH) protocol with a public-private key pair for authentication for Linux instances. For Windows instances, you can use the Remote Desktop Protocol (RDP) client with a username and password.

Yes. You can run a custom startup script as part of the provisioning workflow by including it in the user_data key/value pair of the metadata attribute in the LaunchInstanceDetails object. For more information, see the LaunchInstanceDetails API documentation.

Yes. You can stop your instance without deleting it. The Compute service supports the following actions for an instance:

• Stop: Shuts down the instance. For Standard VM and bare metal instances, billing pauses. For HPC, High I/O bare metal and Dense I/O bare metal and VM instances, billing continues until the instance is terminated.
• Reboot: Restarts the instance (issuing an ACPI reboot).
• Terminate: Shuts down the instance and releases the compute resources. This action can't be reversed. You can keep the instance's boot volume for later reuse. If you choose not to keep the boot volume, it's deallocated and wiped out. Data on the local NVMe SSDs (for storage-optimized instances) is securely wiped. Billing stops for the compute instance.

Customers can consume Oracle Cloud Infrastructure resources without any upfront commitment and pay only for what they use by creating an account at shop.oracle.com. Alternatively, existing customers can contact their sales representative to enable an existing pool of credits, or purchase a new pool, to consume Oracle Cloud Infrastructure resources based on published metered rates.

Oracle Cloud Infrastructure provides you with flexibility, access, and control over your compute instances. You can use a variety of monitoring tools, such as Microsoft SysInternals (sysmon, diskmon, process monitor) and Linux monitoring tools (sysstats, vmstate, iostate), or enterprise management tools such as Oracle Enterprise Manager, to monitor the health of your compute instances. For more information, see the Oracle Enterprise Manager documentation.

You can access Oracle Cloud Infrastructure Compute via the Oracle Cloud Infrastructure Console, REST API, or SDKs. Oracle customers can consume all resources with no upfront commitment and pay only for what they use by creating an account at shop.oracle.com. Alternatively, customers can contact their sales representative to enable an existing pool of credits, or purchase a new pool, and start consuming Oracle Cloud Infrastructure resources based on published metered rates.

After your account is provisioned, see the Getting Started Guide in the service documentation for more information. We have provided a tutorial guiding you through the steps to launch your first instance.

Oracle Cloud Infrastructure will continue to support customers running prior-generation compute instances. However, as we continue to release new instances types, we encourage customers to use the latest offerings to ensure the best performance and pricing. To review prior-generation instances and upgrade recommendations, see the following list.

Bare Metal High I/O (BM.HighIO1.36)

For customers using the BM.HighIO bare metal shape, we recommend upgrading to X7-based Dense I/O virtual machine instances. The X7-based VM.DenseIO2.16 shape offers the newer Intel Skylake processor, including 16 OCPUs, 240 GB of RAM, and 12.8 TB of NVMe storage at a lower price. Additionally, the X7-based VM.DenseIO2.16 shape offers higher network bandwidth. For details, see Compute Shapes in the service documentation. Customers with CPU-intensive workloads can also opt for a higher-CPU offering with the VM.DenseIO2.24 shape, which includes 24 OCPUs, 320 GB of RAM, and 25.6 TB of NVMe storage.

"X5" Series Compute Instance Shapes

Oracle Cloud Infrastructure Compute's autoscaling capability enables you to automatically scale Compute instances for a workload in response to changing performance metrics, instead of manually calling the Compute APIs or using a Terraform script. A guided Console experience and autoscaling APIs help you create autoscaling policies that automatically launch or terminate instances based on metrics emitted by instances in instance pools. As load increases, new instances are dynamically provisioned. And as load decreases, instances are automatically removed.

This capability is available at no additional cost for virtual machine (VM) instances in commercial regions.

Autoscaling acts on instance pools and relies on performance metrics that are collected by the Oracle Cloud Infrastructure Monitoring service. Instance pools allow you to group VM instances together and keep all VM instances in your pool healthy and running. Monitoring lets you capture average CPU and memory utilization metrics from instances in your instance pools.

With autoscaling, you configure thresholds on the aggregate metrics that Monitoring captures from instances in your instance pool. Autoscaling events are triggered when the thresholds are met or exceeded based on your autoscaling policy and rules. Autoscaling responds to a trigger by adding or removing instances from your instance pools. Autoscaling works by changing the size of your instance pool: instance pools are self-managing and will terminate or launch new instances until the pool reaches the target size.

You need to first launch an instance pool using an initial size, and then define an autoscaling configuration for the instance pool. When you define an autoscaling configuration, it sets up Monitoring such that metrics can start flowing from your instances into Monitoring. Monitoring then averages the metrics and sends back the averaged metrics to autoscaling, which triggers events and starts scaling based on the thresholds you set.

An autoscaling configuration defines the information that is needed to configure autoscaling for an instance pool. It contains an autoscaling policy with scaling rules. It also defines a cooldown period between scaling events.

Each instance pool can have one autoscaling configuration.

An autoscaling policy defines the minimum and the maximum number of instances to scale to and the initial size of the instance pool. It also defines the autoscaling policy type. At this time, we support threshold-based autoscaling policies. Each autoscaling policy also contains two autoscaling rules, one rule for scaling in and another rule for scaling out.

You can create one autoscaling policy per autoscaling configuration.

An autoscaling rule defines which metric to use for scaling and the thresholds for scaling in and out. You choose a single metric that is used for both the scale-in rule and the scale-out rule. You also define the number of instances to add or remove with each scaling action.

Autoscaling supports the CPU utilization and memory utilization metrics.

You can add or remove instances in increments of one, up to the maximum instance pool size of 50.

You can scale in and scale out within minutes. You can also control how frequently autoscaling is triggered by adjusting a cooldown period that defines how long to wait between autoscaling actions.

A cooldown period is the minimum amount of time that autoscaling waits before taking another scaling action. It lets the instance pool stabilize at the updated level. The cooldown period starts after the instance pool reaches the target size from the previous scaling event. The minimum cooldown period is five minutes.

Instance pools monitor the life cycle state of instances. If the number of instances in the running or launching state is below the instance pool size, the instance pool creates more instances. If the number of instances in the running or launching state is less than the target size, the instance pool creates more instances.

There are no service limits associated directly with autoscaling. However, instances that are created with autoscaling actions count against your Compute instance limits. When you reach your Compute instance limits, autoscaling is unable to add more instances to your instance pools. See the Service Limits documentation for the default limits for each instance type and instructions on how to request a service limit increase.

You can attach a load balancer working set OCID to your instance pool. After you do this, every time an instance is added to the instance pool, its IP address is also added to the backend set. When the instance reaches a healthy state (the instance is listening on the configured port number), incoming traffic is automatically routed to the new instance. Similarly, every time an instance is removed, the IP address is also removed from the backend set. When removing instances from the load balancer working set, autoscaling waits two minutes before terminating the instance. The two minute delay allows the Load Balancing service to drain connections on the IP address for the instance before the instance is terminated. Any connections that are still active after two minutes are terminated when the instance is shut down.

If the instance configuration was based on an Oracle-provided image released after November 18, 2018, the OracleCloudAgent that emits metrics and works with Monitoring and autoscaling is already installed. You can also manually build, install, and then enable the agent for your custom images.

We support all VM instance shapes including VM.Standard1, VM.DenseIO1, VM.Standard2, VM.DenseIO2, VM.Standard2.E2, VM.GPU2, and VM.GPU3.

We provide images with a variety of Linux distributions such as Oracle Linux, CentOS, and Ubuntu, as well as Microsoft Windows Server. For a complete list and more details, see Oracle-Provided Images help documentation. We support Oracle Linux OSs with Oracle Linux Premier Support included at no additional charge with Oracle Cloud Infrastructure.

You can launch any of the supported OSs on any Oracle Cloud Infrastructure Compute bare metal or virtual machine compute instance shape. You can see a list of all available shapes on the Compute Pricing page.

Yes. Oracle Cloud Infrastructure Compute supports bringing your own OS image. For more information, see Bring Your Own Image help documentation.

Oracle Cloud Infrastructure compute instances running Oracle Linux come included with Oracle Ksplice. Ksplice is an OS-management technology that lets you update important kernel and user-space security components without requiring a reboot.

For more information about downloading patches to your OS instance, see Oracle Ksplice.

Yes. You can create a custom image of your boot disk as a backup or use it as a mechanism to package your preconfigured OS image and use it to launch new compute instances. This is particularly useful when you need to create multiple compute instances with similar configurations. The custom image is instance-type and shape agnostic; it can be used to launch any instance types or shapes: bare metal or virtual machine instances.

After you initiate image creation, the system stops your compute instance to ensure a consistent boot disk image. The time it takes for the image creation process depends on the size of the boot disk. After the image creation is complete, the compute instance automatically restarts. For applications and services that don't automatically start when the instance reboots, you're expected to restart them manually.

Currently, we support the English versions of Windows Server 2008 R2, Windows Server 2012 R2, and Windows Server 2016 as part of the Oracle Cloud Infrastructure image library. Microsoft Windows Server 2019 edition, Oracle Cloud Infrastructure provides support to import your own custom.

The cost of a Microsoft Windows Server license is an additional cost, on top of the underlying compute instance price. You are charged for the compute instance and Windows Server license separately. Your Windows Server license charge starts when your instance is in the "running" state and ends when you terminate the instance. When your instance is in a "stopped state" for Standard VM and Bare Metal instances, billing pauses for both compute instance cost and Microsoft Windows Server operating system License cost. For HPC, High I/O bare metal and Dense I/O bare metal and VM instances, billing continues until the instance is terminated. You can get more information about Microsoft Windows Server pricing from the Compute Pricing page.

Oracle Cloud Infrastructure enables automatic updates on Microsoft Windows Server and uses the default settings for applying Windows patches. You need to update your VCN security list to enable egress traffic for port 80 (http) and port 443 (https) to install patches from Microsoft Windows Update Servers.

Yes. You can create a custom image of your Windows Server instance and use it to launch a new compute instance. We support the creation of "generalized" images (used to create a template or golden image) and "specialized" images (used as backup) for your Windows instance. For more information about these image types, see Creating Windows Custom Images help documentation.

Yes. You can bring your own license (BYOL) for Microsoft applications subject to Microsoft's Terms and Conditions. BYOL on a dedicated host (KVM hypervisor only) is only permitted for Microsoft Windows Server, if the licenses were purchased prior to October 1, 2019. Any licenses purchased or upgraded on or after October 1, 2019 for use on a dedicated host is not permitted on OCI.

The following table shows the Microsoft software for which you can BYOL to Oracle Cloud Infrastructure:

*Subject to Microsoft's Terms and Conditions.

**Application licenses might require License Mobility through Software Assurance when running on Oracle Cloud Infrastructure Windows Server instances. Questions about your licensing rights should be directed to Microsoft or your Microsoft reseller.

***You need appropriate licenses for each user (purchased MSDN subscription).

No. You must bring your own Microsoft Windows image if you bring your Microsoft Windows license.

Oracle provides support for Microsoft Windows Server licensed from Oracle Cloud Infrastructure. All other Microsoft software is supported directly by Microsoft Support.If there are issues that require infrastructure support, Oracle Cloud Infrastructure Support will provide the assistance needed to help you and Microsoft Support resolve the issue.

All third-party images are accessible in the Oracle Cloud Infrastructure Console via the embedded Marketplace. Choose the partner image that you want to launch, and you are guided through the Launch Instance process.

After the partner image is installed, any product issues related to the image are addressed by the partner. For support issues related to installing images, submit a ticket in My Oracle Support.

Yes. From the Create Compute Instance wizard, you can launch prebuilt Oracle enterprise images and solutions that are enabled for Oracle Cloud Infrastructure.

Depending on the Oracle application, there are 30-day evaluation trial terms. At the end of the trial term, a license is required to continue use. Any trial information is detailed in the preinstallation notes in the image selection.

Oracle Cloud Infrastructure provides the following storage options to support different use cases:

• Local storage: Local NVMe-based SSD storage with very high IOPS and low latency is available for the storage-optimized instance type (that is, BM.DenseIO1.36).
• Block volumes: Oracle Cloud Infrastructure Block Volumes provides persistent block storage volumes that can be attached to a compute instance and used like a normal file system for the OS.
• Object storage: Oracle Cloud Infrastructure Object Storage provides highly durable storage for objects such as images, videos, and other unstructured data. Object storage is replicated across availability domains within a region to ensure availability and durability.

Local NVMe SSDs provide very high performance storage and are ideal for the most demanding workloads. For more information about NVMe SSD performance, see the product specification.

After you terminate your compute instance, you have the option to keep its boot volume for later reuse. If you choose not to keep the boot volume, it's deallocated and wiped out. Local NVMe storage for storage-optimized instances is wiped out and deallocated. Any data in the block volumes attached to the instance persists and remains available for later use.

Before terminating the instance, you can create a custom image of the boot disk and use the custom image to launch new instances at a later time.

Bare metal instances are on-demand, bare metal compute resources in the cloud. Unlike virtual machine (VM) instances, bare metal instances are entire physical hosts dedicated to a single customer's use with no hypervisor or Oracle applied software installed. You have full control of the bare metal host's resources, which gives you flexibility, control, and performance without compromise. Bare metal compute instances are ideal for your most demanding applications and workloads.

Oracle Cloud Infrastructure Compute offers a variety of instance types. Each instance type provides a different hardware configuration to support a variety of applications and workloads.

Standard instances can be used for a variety of high-performance and compute-intensive workloads, such as web servers, batch processing, ad serving, and distributed analytics.

Dense I/O instances are configured with 28.8 TB of local NVMe SSD storage and are ideal for extreme transactional workloads that work on large data sets and require low latency and high throughput, such as big data and high-performance computing (HPC) applications.

See the Service Limits help documentation for the default limits for each instance type and instructions on how to request a service limit increase. We are happy to increase the limits for your account as needed.

Currently, Oracle Cloud Infrastructure doesn't provide built-in capability to migrate a bare metal instance to a different bare metal instance type or virtual machine (VM) shape. However, you can perform the migration manually by launching a new compute instance from a boot image that you created (by using the Create Image feature) and subsequently attaching your block volumes to the new instance. If you have data persisted locally, you must copy or replicate the local data manually to the new instance. You can also mitigate changing the shape or scaling for an instance via boot volumes. When you terminate an instance, you can keep its boot volume and launch a new instance with a different shape by using the boot volume that you kept from the original instance.

Yes. Bare metal instances are dedicated physical hosts with no hypervisor installed on them. You have access to the host and can install your own type 2 hypervisor (hosted hypervisor) such as KVM or VirtualBox to run any version of OS supported by the hypervisors, subject to your existing licensing terms.

Oracle Cloud Infrastructure Compute VMs offer smaller compute shapes than bare metal instances. VMs use the same cloud-optimized hardware and networking infrastructure as bare metal instances to deliver compute shape flexibility and performance for your changing application needs.

Oracle Cloud Infrastructure Compute offers different VM instance shapes to meet the varying needs of the applications that you build. New instance shapes are added frequently. For details, see the Compute Pricing page.

See the Service Limits help documentation for the default limits and instructions on how to request a service limit increase. We are happy to increase the limits for your account as needed.

Each VM shape has a different number of processor cores, amount of memory, and disk configuration, so you can easily select a size that matches the needs of your application.

There are two types of VM shapes.

• Standard VM compute shapes are great for general purpose workloads, providing a balance of cores, memory, and network resources.
• DenseIO VM compute shapes are better suited for resource-intensive workloads, such as large databases, and big data applications like Apache Spark and Hadoop.

You can use the compute Console, API or CLI to change the shape of a VM after it's created. You can choose any VM shape that is compatible with the existing operating system image. The shape change will require a reboot but will preserve all instance properties. See the technical documentation for more information about how to change the shape of an instance.

You can also migrate your VM manually by launching a new compute instance from a boot image that you created (by using the Create Image feature) and subsequently attaching your block volumes to the new instance. If you have data persisted locally, you must copy or replicate the local data manually to the new instance.

Oracle Cloud Infrastructure Compute dedicated virtual machine hosts enable you to run VM compute instances on a physical server that is dedicated to your Oracle Cloud Infrastructure tenancy. No VMs from any other tenancy or Oracle customer will run on this dedicated host. You can then set an optional parameter when launching a VM compute instance to place it on your Dedicated Virtual Machine Host. These dedicated hosts are "single tenant" - i.e., dedicated exclusively to your tenancy. VM characteristics, such as number of OCPUs and amount of memory, and performance are the same as when running on regular shared hardware in Oracle Cloud Infrastructure.

Oracle Cloud Infrastructure Compute DVH is currently offered in a DVH.Standard2.52 shape. This shape allows you to launch any VM.Standard2 instance on the DVH. New DVH shapes will be added over time.

You are charged for the dedicated virtual machine host when it is in the running state. You are not charged for the individual compute VM instances launched on the host, however, you will still be charged per OCPU hour for a Windows Server license when you run a VM instance using a Windows Server image. Billing stops when you terminate the dedicated host. For the DVH.Standard2.52 host, you are charged for a 52 OCPU BM.Standard2.52 instance once the dedicated virtual machine host starts running. On your bill, you will see a charge for 52 OCPUs under "Oracle Cloud Infrastructure Compute Bare Metal Standard - X7". You can at maximum use 48 OCPUs for running your VMs in a DVH.Standard2.52.

The only supported actions with a dedicated virtual machine host are to launch, move compartments and terminate. However, you are not allowed to terminate the dedicated virtual machine host if there are VM instances still running on it. You are required to terminate all VM instances before you can terminate the dedicated virtual machine host.

Oracle Cloud Infrastructure provides several options for sending feedback or getting support from the community and from Oracle. You can find vibrant and active community support in the Oracle Forum and regular updates via the Oracle Cloud Infrastructure Blog.

You can also get support for debugging and troubleshooting by submitting a service request via My Oracle Support.