Network-optimized machine family for Compute Engine

The network-optimized machine family is suitable for workloads that run network-intensive applications, such as telecommunications, network appliances, and financial services industry (FSI) applications. These machine series offer the highest network and block storage performance within Compute Engine.

Machine series Workloads
M4N
  • High-performance vector databases
  • Retrieval-augmented generation (RAG) data layers
  • Massive in-memory context caching
  • Real time semantic search

M4N machine series

M4N is engineered to deliver the highest I/O performance available, reaching up to 12,500 MiB/s of combined host storage performance. This new level of performance opens up new possibilities for I/O-intensive workloads that require extreme bandwidth for data access and processing.

M4N instances are powered by the Intel Emerald Rapids processors and Titanium offload processors. This machine series is well-suited for large in-memory databases such as SAP HANA, as well as online analytical processing (OLAP) and in-memory data analytics workloads. The M4N machine series offers a lower cost per transaction for Oracle Database workloads and the best TCO when compared to any other compute instance for databases that require licencing per core.

M4N machine types allow you to provision up to 112 vCPUs and up to 2.9 TB of RAM. M4N instances use only NVMe for storage, and support Hyperdisk Balanced and Hyperdisk Extreme. M4N instances use only gVNIC for networking. VirtIO-net and SCSI interfaces are not supported.

M4N offers the following features:

  • Superior VM-to-VM network bandwidth: Achieves up to 200 Gbps of VM-to-VM network bandwidth and supports up to 50 Gbps single-flow bandwidth between M4N instances routed within the same VPC network.
  • Enhanced VM to Internet performance: Internet egress network bandwidth can reach up to 100 Gbps. Internet egress packet processing performance can scale up to 24 million packets per second (MPPS).
  • Industry-leading packet processing: Up to 48 MPPS of sustained packet processing performance (measured using DPDK Pktgen).
  • Optimized I/O for smaller shapes: Up to 100 Gbps of network bandwidth specifically for machine types with 16 to 56 vCPUs without requiring premium infrastructure add-ons (such as per VM Tier_1 networking performance).
  • Enhanced out-of-the-box networking: gVNIC-type network interfaces on M4N now start with 64 Tx/Rx queues by default.
  • Titanium-powered efficiency with Dual-NIC Titanium architecture: Uses a 200G Titanium network adapter to fully offload network and storage management, ensuring your applications run with maximum performance and predictability.
  • Leading Block Storage performance and options: Supports Hyperdisk Balanced and Hyperdisk Extreme block storage options. M4N with Hyperdisk Extreme provides the low-latency, high-speed data access that modern databases and enterprise AI applications need, with up to 12.5 GiB/s of block storage throughput and 500,000 IOPS.

M4N Limitations

The M4N machine series has the following restrictions:

  • M4N machine types are only available in a predefined machine type. Custom machine types aren't available.
  • You can't use GPUs with M4N instances.
  • M4N instances are available in only select zones and regions.

M4N machine types

Machine types vCPUs1 Memory (GB) Local SSD Maximum internal network bandwidth
(Gbps)2		 Maximum external network bandwidth
(Gbps)2		 NUMA domains
m4n-hypermem-16 16 248 Not available Up to 50 Up to 25 1
m4n-hypermem-32 32 496 Not available 57 28 1
m4n-hypermem-64 64 992 Not available 114 57 2
m4n-megamem-28 28 372 Not available 50 25 1
m4n-megamem-56 56 744 Not available 100 50 1
m4n-megamem-112 112 1,488 Not available 200 100 2
m4n-ultramem-28 28 744 Not available 50 25 1
m4n-ultramem-56 56 1,488 Not available 100 50 2
m4n-ultramem-112 112 2,976 Not available 200 100 4

1 Each virtual CPU (vCPU) is implemented as a single hardware multithread, and two vCPUs share each physical CPU core by default.
2 Maximum egress bandwidth cannot exceed the number given. Actual egress bandwidth depends on the destination IP address and other factors. See Network bandwidth.

Supported disk types for M4N

M4N instances can use the following block storage types:

  • Hyperdisk Balanced (hyperdisk-balanced)
  • Hyperdisk Extreme (hyperdisk-extreme)

Disk and capacity limits

The following restrictions apply:

  • The number of Hyperdisk volumes can't exceed 64 per VM.
  • The maximum total disk capacity across all disks can't exceed 512 TiB.

For details about the capacity limits, see Hyperdisk capacity limits per VM.

M4N storage limits are described in the following table:

Maximum number of disks
Machine types Hyperdisk
per VM		 Hyperdisk Balanced Hyperdisk Throughput Hyperdisk Extreme
m4n-hypermem-16 16 16 0 8
m4n-hypermem-32 32 32 0 8
m4n-hypermem-64 32 32 0 8
m4n-megamem-28 32 32 0 8
m4n-megamem-56 32 32 0 8
m4n-megamem-112 64 64 0 8
m4n-ultramem-28 32 32 0 8
m4n-ultramem-56 32 32 0 8
m4n-ultramem-112 64 64 0 8

Network support for M4N instances

M4N instances require gVNIC network interfaces. M4N instances support up to 100 Gbps network bandwidth for standard networking and don't support per VM Tier_1 networking performance.

Before migrating to M4N or creating M4N VM instances, make sure that the operating system image that you use supports the gVNIC driver for VM instances. These images include an updated gVNIC driver, even if the guest OS shows the gve driver version as 1.0.0. If your M4N VM is using an operating system with an older version of gVNIC driver, this is still supported but the VM might experience suboptimal performance such as less network bandwidth or higher latency.

If you use a custom OS image to create a M4N VM, you can manually install the most recent gVNIC driver. The gVNIC driver version v1.4.2 or later is recommended for use with M4N VMs. Google recommends using the latest gVNIC driver version to benefit from additional features and bug fixes.

Maintenance experience for M4N instances

During the lifecycle of a Compute Engine instance, the host machine that your instance runs on undergoes multiple host events. A host event can include the regular maintenance of Compute Engine infrastructure, or in rare cases, a host error. Compute Engine also applies some non-disruptive lightweight upgrades for the hypervisor and network in the background.

The M4N machine series offers the following features related to host maintenance:

Machine type Typical scheduled maintenance event frequency Maintenance behavior Advanced notification On-demand maintenance Simulate maintenance
All machine types Monthly Live migrate 7 days Yes Yes

The maintenance frequencies shown in the previous table are approximations, not guarantees. Compute Engine might occasionally perform maintenance more frequently.

What's next