This document describes the features of Google Cloud Hyperdisk.Hyperdisk is the fastest and most efficient durable disk forCompute Engine. If you need boot or data disks for your computeinstances—virtual machine (VM) instances, containers, and bare metalinstances—then Google recommends using Hyperdisk.

For information about the other block storage options inCompute Engine, seeChoose a disk type.

To create a new Hyperdisk volume, seeCreate a Hyperdisk volume.

Hyperdisk features

With Hyperdisk you can provision, manage, and scale yourCompute Engine workloads without the cost and complexity of a typicalon-premises storage area network (SAN).

Hyperdisk volumes have the following features:

• Function as physical disks: you can use a Hyperdiskvolume with a compute instance as if it were a physical disk attached to theinstance. When you read to or write from a Hyperdisk volume,data is transmitted over the network.

• Higher performance: Hyperdisk offers higher IOPS andthroughput than Persistent Disk by leveraging Google's Titaniumstorage offload technology.

• Customizable performance: you can choose the performance—IOPS orthroughput—of each Hyperdisk volume. You can alsoincrease or decrease a Hyperdisk volume's performance whileit's in use.

• Support for high availability: in the unlikely event of a zonal orregional outage, you can ensure high availability for your data by enablingone or both of the following features:

  • To protect your data in case of a zonal outage, useHyperdisk Balanced High Availability.Data on Hyperdisk Balanced High Availability volumes is synchronously replicated across two zoneswithin the same region to protect against up to one zonal outage.

  • To protect your data from a regional outage, maintain a replica of yourdata in another region by usingAsynchronous Replication.When you enable Asynchronous Replication for a disk, data in one region iscontinuously copied to a replica in a secondary region. If a regionaloutage occurs, you canfailoveryour data to a secondary region. Asynchronous Replication is available forHyperdisk Balanced, Hyperdisk Balanced High Availability, and Hyperdisk Extreme volumes.

• Portability: you can change the compute instance that aHyperdisk volume is attached to.

• Shareable between VMs: for high availability workloads, certainHyperdisk types can be shared by multiple VMs. Each VM hassimultaneous read-write or read-only access to the volume.

• Support for pooled capacity and performance: to simplify planning,avoid overprovisioning storage, and reduce costs, you can purchaseHyperdisk storage and performance in bulk by usingHyperdisk Storage Pools.

Choose a Hyperdisk type for your workload

To add Hyperdisk volumes to your workloads, you must choose aHyperdisk type. Each Hyperdisk type is designedand optimized for a specific type of workload.The following is a list of the available Hyperdisk types.

• Hyperdisk Balanced
• Hyperdisk Balanced High Availability
• Hyperdisk Extreme
• Hyperdisk Throughput
• Hyperdisk ML

For most workloads, we recommend Hyperdisk Balanced.

To select a Hyperdisk type, compare your workload's type and itsperformance requirements with the information in the following table. Fordetailed information about a specific Hyperdisk type, see thelinked page in theRecommended Hyperdisk type column.

¹ You can't specify a throughput level for Hyperdisk Extreme volumes. The provisioned throughput is based on the IOPS level you specify.
² You can't specify an IOPS level for Hyperdisk Throughput and Hyperdisk ML volumes. The provisioned IOPS is based on the throughput level you specify.

Hyperdisk size limits

The following table lists the size limits for each Hyperdisk type.

However, the size of a Hyperdisk volume that's attached to an instancecan't exceed the instance'sHyperdisk limits.

Hyperdisk performance

The following is a summary of key Hyperdisk performance concepts:

• You can configure the performance (IOPS and/or throughput) limit and size ofeach Hyperdisk volume. You can also increase or decrease aHyperdisk volume's performance without changing its size.
• The performance limit you specify is referred to as theprovisioned performance. The provisioned performance isn't the expectedperformance, rather, it's the maximum performance the disk can achieve.
• The actual performance for a Hyperdisk volume is the observedperformance while the volume is in use.
• For a Hyperdisk volume to reach its provisioned performance,you must attach it to a compute instance that supports the same levelof performance or higher.

For a discussion of how Hyperdisk performance works, seeAbout Hyperdisk performance.For performance limits for each Hyperdisk type, seeHyperdisk performance limits.

Latency

Each Hyperdisk type has different latency profiles. Googlerecommends comparing Hyperdisk Throughput to the latency of a hard disk drive. You cancompare the latency for Hyperdisk Balanced, Hyperdisk Balanced High Availability, Hyperdisk Extreme, and Hyperdisk ML to thelatency of enterprise SSDs.

Hyperdisk Balanced and Hyperdisk Extreme offer sub-millisecond latency.

Machine series support for Hyperdisk

This section lists themachine seriesthat each Hyperdisk type supports.If a machine series doesn't support Hyperdisk, use Persistent Disk.

Select one or more machine series to see the supported Hyperdisk types.

Restrictions for machine series support

This section lists the restrictions that apply to the machine seriesthat each Hyperdisk type supports.

• For Hyperdisk Extreme, the following restrictions apply:
  • C3 machine type require at least 88 vCPUs.
  • C3D machine types require at least 60 vCPUs.
  • C4 and G4 machine types require at least 96 vCPUs.
  • M1 machine types require at least 80 vCPUs.
  • C4A, C4D, M3, and M4 machine types require at least 64 vCPUs.
  • N2 machine types require at least 80 vCPUs; Custom N2 machine types aren't supported
• For Hyperdisk Throughput, the following restrictions apply:
  • You can't use Hyperdisk Throughput with bare metal instances.
  • Additional limitations apply to attaching Hyperdisk Throughput volumes to certain machine series, including C4, C4A, and N4. For more information, see Limitations for attaching Hyperdisk Throughput volumes to 4th-generation instances.

Regional availability for Hyperdisk

Some Hyperdisk types are available in allregions and zones, whileothers are available only in specific locations. The following table summarizesregional availability for each Hyperdisk type.

Share Hyperdisk volumes between VMs

You can share a Hyperdisk volume between multiple VMs bysimultaneously attaching the same volume to multiple VMs.

The following scenarios are supported:

• Concurrent read-write access to a single volume from multiple VMs.Recommended for clustered file systems and highly available workloads likeSQL Server Failover Cluster Infrastructure. Supported for the followingHyperdisk types:

  • Hyperdisk Balanced
  • Hyperdisk Extreme
  • Hyperdisk Balanced High Availability

• Concurrent read-only access to a single volume from multiple VMs.This is more cost effective than having multiple disks with the same data.Recommended for accelerator-optimized machine learning workloads.Supported for Hyperdisk ML volumes.

To learn about disk sharing, seeShare a disk between VMs.

High availability and disaster recovery protection for Hyperdisk volumes

You can protect your data in the rare event of a zonal or regional outage byenabling replication, that is, maintaining a copy of the data in another zone orregion.

Cross-zonal synchronous replication

To replicate data to another zone within the same region, you must use Hyperdisk Balanced High Availabilityvolumes. Hyperdisk Balanced High Availability is the only supported Hyperdisk type forzonal replication.

For more information, seeAbout synchronous disk replication.

Cross-regional asynchronous replication

You can protect your data in the unlikely event of a regional outage by enabling Asynchronous Replication. Asynchronous Replication maintains a copy ofthe data on your volume in another region. For example, to protect aHyperdisk volume inus-west1,you can use Asynchronous Replication to replicate the volume to a secondary volume intheus-east4 region. If the volume inus-west1became unavailable, then you could use the secondary volume inus-east4.

You can use Asynchronous Replication with the following Hyperdisk types:

• Hyperdisk Balanced
• Hyperdisk Extreme
• Hyperdisk Balanced High Availability

To learn more about cross-regional replication, seeAsynchronous Replication.

Encryption for Hyperdisk volumes

By default, Compute Engine protects your Hyperdisk volumes withGoogle-owned and Google-managed encryption keys. You can also encrypt yourHyperdisk volumes with customer-managed encryption keys (CMEK).

For more information, seeAbout disk encryption.

Confidential Computing with Hyperdisk volumes

You can add hardware-based encryption to a Hyperdisk Balanced disk by enablingConfidential mode for the disk when you create it.You can use Confidential mode only with Hyperdisk Balanced disks that are attached toConfidential VMs.

For more information, seeConfidential mode for Hyperdisk Balanced volumes.

Durability of Hyperdisk

Compute Engine distributes the data on Hyperdisk volumesacross several physical disks to ensure durability and optimize performance.

Disk durability represents the probability of data loss, by design, for atypical disk in a typical year. Hyperdisk data loss events areextremely rare and have historically been the result of coordinated hardwarefailures, software bugs, or a combination of the two. Google takes many steps tomitigate the industry-wide risk of silent data corruption.

Durability is calculated with a set of assumptions about hardware failures,the likelihood of catastrophic events, isolation practices and engineeringprocesses in Google data centers, and the internal encodings used by each disktype.

Human error by a Google Cloud customer, such as when a customer accidentallydeletes a disk, is outside the scope of Hyperdisk durability.

The table below shows durability for each disk type's design. 99.999% durabilitymeans that with 1,000 Hyperdisk volumes, you would likely go ahundred years without losing a single one.

1. Create a pool: create a pool withthe estimated capacity and performance forall workloads in a project in a specific zone. Each poolcan contain one type of Hyperdisk, such as Hyperdisk Throughput.
2. Create disks in the pool:create disks in this pool and attach themto existing instances. You can also create a disk in the poolwhen you create a new instance.

You can use Hyperdisk Balanced or Hyperdisk Throughput volumes with Hyperdisk pools.Compute Engine offers the following types of pools:

• Hyperdisk Storage Pools for workloads that require up to 5 PiB ofcapacity per project per zone.
• Hyperdisk Exapools for workloads that require up to5 EiB of capacity per project per zone.

For more information about pools, seeAbout Hyperdisk pools.

Pricing

You are billed for the total provisioned capacity of yourHyperdisk volumes until you delete them. Charges incur even ifthe volume isn't attached to any instances or if the instance is suspended orstopped. You are charged per GiB per month. Additionally, you are billed for thefollowing:

• Hyperdisk Balanced charges a monthly rate for the provisioned IOPS and provisionedthroughput (in MiB/s) in excess of the baseline values of3,000 IOPS and140 MiB/s throughput.
• Hyperdisk Extreme charges a monthly rate based on the provisioned IOPS.
• Hyperdisk ML charges a monthly rate based on the provisioned throughput(in MiB/s).There is no additional charge for attaching multiple VMs to a singleHyperdisk ML volume.
• Hyperdisk Throughput charges a monthly rate based on the provisioned throughput(in MiB/s).

Because the data for regional disks is written to two locations,the cost of Hyperdisk Balanced High Availability storage is twice the cost of Hyperdisk Balanced storage.

For more pricing information, seeDisk pricing.

Hyperdisk and committed use discounts

Hyperdisk volumes are not eligible for:

• Resource-based committed use discounts (CUDs)
• Sustained use discounts (SUDs)

Hyperdisk and preemptible VM instances

Hyperdisk can be used with Spot VMs (orpreemptible VMs). However, there are no discounted spot prices forHyperdisk.

Limitations for Hyperdisk

• You can't create amachine image from aHyperdisk volume.
• You can'tcreate an imagefrom a Hyperdisk Extreme, Hyperdisk Throughput, or Hyperdisk Balanced High Availability volume.
• You can't create an instant snapshot from a Hyperdisk ML or Hyperdisk Throughput volume.
• Hyperdisk Extreme, Hyperdisk ML and Hyperdisk Throughput volumes can't be used as boot disks.
• You can attach a Hyperdisk ML volume to up to 100 VMs at most once every30 seconds.
• You can't create a Hyperdisk ML disk in read-write mode from a snapshot or adisk image. You must create the disk in read-only mode.
• If you enable read-only mode for a Hyperdisk ML volume, you can't re-enableread-write mode.
• If you create a volume in multi-writer mode, seeadditional limitations.
• If you create a Hyperdisk Balanced volume in Confidential mode, seeadditional limitations.
• Confidential VMs with AMD SEV on C3D machine types don't supportHyperdisk Balanced and Hyperdisk Throughput.

What's next?

• Learn how tocreate a Hyperdisk volume.
• Learn how toclone a Hyperdisk volume.
• Learn aboutsynchronous disk replicationwith Hyperdisk Balanced High Availability.
• Learn aboutHyperdisk pools.
• ReviewDisk pricing information.
• Learn how tooptimize performance of Hyperdisk.