Cloud Load Balancing overview

A load balancer distributes user traffic across multiple instances of your applications. By spreading the load, load balancing reduces the risk that your applications experience performance issues. Google's Cloud Load Balancing is built on reliable, high-performing technologies such as Maglev, Andromeda, Google Front Ends, and Envoy—the same technologies that power Google's own products.

Cloud Load Balancing offers a comprehensive portfolio of application and network load balancers. Use our global proxy load balancers to distribute millions of requests per second among backends in multiple regions with our Google Front End fleet in over 80 distinct locations worldwide—all with a single, anycast IP address. Implement strong jurisdictional control with our regional proxy load balancers, keeping your backends and proxies in a region of your choice without worrying about TLS/SSL offload. Use our passthrough load balancers to quickly route multiple protocols to backends with the high performance of direct server return (DSR).

Key features of Cloud Load Balancing

Cloud Load Balancing offers the following load balancer features:

• Single anycast IP address. With Cloud Load Balancing, a single anycast IP address is the frontend for all of your backend instances in regions around the world. It provides cross-region load balancing, including automatic multi-region failover, which moves traffic to failover backends if your primary backends become unhealthy. Cloud Load Balancing reacts instantaneously to changes in users, traffic, network, backend health, and other related conditions.

• Seamless autoscaling. Cloud Load Balancing can scale as your users and traffic grow, including easily handling huge, unexpected, and instantaneous spikes by diverting traffic to other regions in the world that can take traffic. Autoscaling does not require pre-warming: you can scale from zero to full traffic in a matter of seconds.

• Software-defined load balancing. Cloud Load Balancing is a fully distributed, software-defined, managed service for all your traffic. It is not an instance-based or device-based solution, so you won't be locked into a physical load-balancing infrastructure or face the high availability, scale, and management challenges inherent in instance-based load balancers.

• Layer 4 and Layer 7 load balancing. Use Layer 4-based load balancing to direct traffic based on data from network and transport layer protocols such as TCP, UDP, ESP, GRE, ICMP, and ICMPv6 . Use Layer 7-based load balancing to add request routing decisions based on attributes, such as the HTTP header and the uniform resource identifier.

• External and internal load balancing. Defines whether the load balancer can be used for external or internal access. An external load balancer accepts traffic from the internet, whereas an internal load balancer only accepts RFC 1918 traffic. You can use external load balancing when your users reach your applications from the internet. You can use internal load balancing when your clients are inside of Google Cloud. To learn more, see external versus internal load balancing.

• Global and regional load balancing. Defines the scope of the load balancer. A global load balancer supports backends in multiple regions, whereas a regional load balancer supports backends in a single region. Even though the IP address of a regional load balancer is located in one region, a regional load balancer is globally accessible. You can distribute your backends in single or multiple regions to terminate connections close to your users and to meet your high availability requirements. To learn more, see global versus regional load balancing.

• Routing of traffic in Premium Tier and Standard Tier. The load balancing services in Google Cloud come in different flavors depending on the network tier you choose, that is, Premium Tier or Standard Tier, with the former being more expensive than the latter. The Premium Tier leverages Google's high-quality global backbone whereas the Standard Tier uses the public internet to route traffic across the network. The network tier you choose depends on whether you prioritize cost or performance for your enterprise workload. Some load balancing services are only available in Premium Tier and not the Standard Tier. To learn more, see Premium versus Standard Network Service Tiers.

• Advanced feature support. Cloud Load Balancing supports features such as IPv6 load balancing, source IP-based traffic steering, weighted load balancing, WebSockets, user-defined request headers, and protocol forwarding for private virtual IP addresses (VIPs).

  It also includes the following integrations:

  • Integration with Cloud CDN for cached content delivery. Cloud CDN is supported with the global external Application Load Balancer and the classic Application Load Balancer.
  • Integration with Google Cloud Armor to protect your infrastructure from distributed denial-of-service (DDoS) attacks and other targeted application attacks. Always-on DDoS protection is available for the global external Application Load Balancer, the classic Application Load Balancer, the external proxy Network Load Balancer, and the external passthrough Network Load Balancer. Additionally, Google Cloud Armor supports advanced network DDoS protection only for external passthrough Network Load Balancers. For more information, see Configure advanced network DDoS protection.

Types of Google Cloud load balancers

Cloud Load Balancing offers two types of load balancers: Application Load Balancers and Network Load Balancers. You'd choose an Application Load Balancer when you need a Layer 7 load balancer for your applications with HTTP(S) traffic. You'd choose a Network Load Balancer when you need a Layer 4 load balancer that supports TLS offloading (with a proxy load balancer) or you need support for IP protocols such as UDP, ESP, and ICMP (with a passthrough load balancer).

The following table provides a high-level overview of the different types of Google Cloud load balancers categorized by the OSI layer on which they operate and whether they are used for external or internal access.

lan Cloud Load Balancing	External (Accepts internet traffic)	Internal (Accepts only RFC 1918 traffic)
Application Load Balancers HTTPS Layer 7 load balancing	global external regional external classic	cross-region internal regional internal
Network Load Balancers TCP/SSL/Other Layer 4 load balancing
	Proxy Network Load Balancers
	global external regional external classic	cross-region internal regional internal
	Passthrough Network Load Balancers
	regional external	regional internal

Application Load Balancers

Application Load Balancers are proxy-based Layer 7 load balancers that enable you to run and scale your services behind an anycast IP address. The Application Load Balancer distributes HTTP and HTTPS traffic to backends hosted on a variety of Google Cloud platforms—such as Compute Engine and Google Kubernetes Engine (GKE)—as well as external backends outside Google Cloud.

The following diagram provides a high-level overview of the different types of Application Load Balancers that can be deployed externally or internally depending on whether your application is internet-facing or internal.

External Application Load Balancers are implemented as managed services either on Google Front Ends (GFEs) or Envoy proxies. Clients can connect to these load balancers from anywhere on the internet. Note the following:

• These load balancers can be deployed in the following modes: global, regional, or classic.
• Global external Application Load Balancers support backends in multiple regions.
• Regional external Application Load Balancers support backends in a single region only.
• Classic Application Load Balancers are global in Premium Tier. In Standard Tier, they can distribute traffic to backends in a single region only.
• Application Load Balancers use the open source Envoy proxy to enable advanced traffic management capabilities.

Internal Application Load Balancers are built on the Andromeda network virtualization stack and the open source Envoy proxy. This load balancer provides internal proxy-based load balancing of Layer 7 application data. The load balancer uses an internal IP address that is accessible only to clients in the same VPC network or clients connected to your VPC network. Note the following:

• These load balancers can be deployed in the following modes: regional or cross-region.
• Regional internal Application Load Balancers support backends only in a single region.
• Cross-region internal Application Load Balancers support backends in multiple regions and are always globally accessible. Clients from any Google Cloud region can send traffic to the load balancer.

To learn more about Application Load Balancers, see Application Load Balancer overview.

Network Load Balancers

Network Load Balancers are Layer 4 load balancers that can handle TCP, UDP, or other IP protocol traffic. These load balancers are available as either proxy load balancers or passthrough load balancers. You can pick a load balancer depending on the needs of your application and the type of traffic that it needs to handle. Choose a proxy Network Load Balancer if you want to configure a reverse proxy load balancer with support for advanced traffic controls and backends on-premises and in other cloud environments. Choose a passthrough Network Load Balancer if you want to preserve the source IP address of the client packets, you prefer direct server return for responses, or you want to handle a variety of IP protocols such as TCP, UDP, ESP, GRE, ICMP, and ICMPv6 .

Proxy Network Load Balancers

Proxy Network Load Balancers are Layer 4 reverse proxy load balancers that distribute TCP traffic to virtual machine (VM) instances in your Google Cloud VPC network. Traffic is terminated at the load balancing layer and then forwarded to the closest available backend by using TCP.

The following diagram provides a high-level overview of the different types of proxy Network Load Balancers that can be deployed externally or internally depending on whether your application is internet-facing or internal.

External proxy Network Load Balancers are Layer 4 load balancers that distribute traffic that comes from the internet to backends in your Google Cloud VPC network, on-premises, or in other cloud environments. These load balancers are built on either Google Front Ends (GFEs) or Envoy proxies.

These load balancers can be deployed in the following modes: global, regional, or classic.

• Global external proxy Network Load Balancers support backends in multiple regions.
• Regional external proxy Network Load Balancers support backends in a single region.
• Classic proxy Network Load Balancers are global in Premium Tier. In Standard Tier, they can distribute traffic to backends in a single region only.

Internal proxy Network Load Balancers are Envoy proxy-based regional Layer 4 load balancers that enable you to run and scale your TCP service traffic behind an internal IP address that is accessible only to clients in the same VPC network or clients connected to your VPC network.

These load balancers can be deployed in one of the following modes: regional or cross-region.

• Regional internal proxy Network Load Balancers support backends in a single region only.
• Cross-region internal proxy Network Load Balancers support backends in multiple regions and are always globally accessible. Clients from any Google Cloud region can send traffic to the load balancer.

To lean more about proxy Network Load Balancers, see proxy Network Load Balancer overview.

Passthrough Network Load Balancers

Passthrough Network Load Balancers are Layer 4 regional, passthrough load balancers. These load balancers distribute traffic among backends in the same region as the load balancer. They are implemented by using Andromeda virtual networking and Google Maglev.

As the name suggests, these load balancers are not proxies. Load-balanced packets are received by backend VMs with the packet's source and destination IP addresses, protocol, and, if the protocol is port-based, the source and destination ports unchanged. Load-balanced connections are terminated at the backends. Responses from the backend VMs go directly to the clients, not back through the load balancer. The industry term for this is direct server return (DSR).

These load balancers, as depicted in the following image, are deployed in two modes, depending on whether the load balancer is internet-facing or internal.

• External passthrough Network Load Balancers are built on Maglev. Clients can connect to these load balancers from anywhere on the internet regardless of their Network Service Tiers. The load balancer can also receive traffic from Google Cloud VMs with external IP addresses or from Google Cloud VMs that have internet access through Cloud NAT or instance-based NAT.

  Backends for external passthrough Network Load Balancers can be deployed using either a backend service or a target pool. For new deployments, we recommend using backend services.

• Internal passthrough Network Load Balancers are built on the Andromeda network virtualization stack. An internal passthrough Network Load Balancer lets you to load balance TCP/UDP traffic behind an internal load-balancing IP address that is accessible only to systems in the same VPC network or systems connected to your VPC network. This load balancer can only be configured in Premium Tier.

To learn more about passthrough Network Load Balancers, see passthrough Network Load Balancer.

Load balancer components

A load balancer is a system composed of multiple interacting components. There is no single API resource that represents a load balancer. Instead, multiple components work together to distribute incoming traffic across multiple backends.

The following diagram shows the core components of an Application Load Balancer, a proxy Network Load Balancer, and a passthrough Network Load Balancer.

The information that follows is a high-level overview of the key components of a load balancer, starting from the point where traffic reaches the load balancer to the stage where it is routed to the backend resource. For a deeper understanding of each load balancer component, refer to the page linked in each section.

Forwarding rule

A forwarding rule specifies an IP address, an IP protocol, and one or more ports on which the load balancer accepts traffic. The forwarding rule and its attached IP address represent the frontend of a Google Cloud load balancer.

For more information, see Forwarding rules overview.

Target proxy

Target proxies terminate incoming connections from clients and create new connections from the load balancer to the backends.

• The first connection originates from the client and is terminated at the target proxy of the load balancer.

• The second connection starts at the target proxy and ends at the backend instance, which handles the client request.

The first connection is terminated either in a Google Front End (GFE) or in a specially designated subnet known as the proxy-only subnet, which is reserved exclusively for Envoy proxies. To know whether a load balancer is GFE-based or Envoy-based, see the table in the Underlying technologies of Google Cloud load balancers section of this document.

Target proxies are used only by proxy-based load balancers such as Application Load Balancers and proxy Network Load Balancers. For these types of load balancers, responses from the backend instances are sent back to the target proxy rather than directly to the client.

For more information, see Target proxies.

Proxy-only subnet

A proxy-only subnet provides a pool of IP addresses that are reserved exclusively for Envoy proxies used by Google Cloud load balancers. The proxies terminate incoming connections and then create new connections to the backend.

For more information, see Proxy-only subnets for Envoy-based load balancers

SSL certificates

Also known as Transport Layer Security (TLS) certificates, SSL certificates facilitate secure communication between clients and load balancers. Proxy-based load balancers whose forwarding rules reference a target HTTPS proxy or a target SSL proxy require a private key and SSL certificate as part of the load balancer's target proxy configuration.

For more information, see SSL certificates.

URL map

Upon terminating the connection, the target HTTP(S) proxy uses the URL map to decide where to route the new request (the second connection as stated in the Target proxy section). The request is routed to either the backend service or the backend bucket. URL maps are only used by Application Load Balancers. As Application Load Balancers operate at Layer 7 of the OSI model, they can make routing decisions based on HTTP attributes, such as domain name, request path, and query parameters.

For more information, see URL maps.

Backend service

A backend service defines how your load balancer distributes traffic. The backend service configuration contains a set of values, such as the protocol used to connect to backends, various distribution and session settings, health checks, and timeouts.

These settings provide fine-grained control over how your load balancer behaves and let you direct traffic to the correct backends, which can be either VM instance groups or network endpoint groups (NEGs).

For more information, see Backend services overview.

Backend bucket

If your workload serves static content using the HTTP(S) protocol, you can use a Cloud Storage bucket to store static content, and then use a backend bucket to route requests to it.

Health checks

When you configure a load balancer's backend service, you need to specify one or more health checks for its backends. A health check, as the name suggests, determines whether the backend instances of the load balancer are healthy. This determination is based on the ability of the backend to respond to incoming traffic. The traffic that a backend needs to respond to depends of the type of the load balancer. You can create health checks using both Layer 7 and Layer 4 protocols to monitor load balanced instances.

Firewall rules

For health checks to work, you must create ingress allow firewall rules that allow health check probes to reach your backends.

Load balancers based on Google Front Ends require an ingress allow firewall rule that permits traffic from the CIDRs of Google Front End to connect to your backends. Load balancers based on the open source Envoy proxy require an ingress allow firewall rule that permits traffic from the proxy-only subnet to reach the backend instances.

For more information, see Firewall rules.

Backends

Backends are the final destination of load-balanced traffic.

Different load balancers support different types of backends. When you add a backend to the backend service, you specify a balancing mode that evaluates the backend's capacity to handle new requests and determines how traffic is distributed among the backends.

For more information, see Backends.

Underlying technologies of Google Cloud load balancers

The following table lists the underlying technology upon which each Google Cloud load balancer is built.

• Google Front Ends (GFEs) are software-defined, distributed systems that are located in Google points of presence (PoPs) and perform global load balancing in conjunction with other systems and control planes.
• Andromeda is Google Cloud's software-defined network virtualization stack.
• Maglev is a distributed system for Network Load Balancing.
• Envoy is an open source edge and service proxy, designed for cloud-native applications.

Load balancer	Technology
Global external Application Load Balancer	Envoy-based Google Front-End (GFE)
Classic Application Load Balancer	GFE
Regional external Application Load Balancer	Envoy
Cross-region internal Application Load Balancer	Envoy
Regional internal Application Load Balancer	Envoy
Global external proxy Network Load Balancer	Envoy-based GFE
Classic proxy Network Load Balancer	GFE
Regional external proxy Network Load Balancer	Envoy
Regional internal proxy Network Load Balancer	Envoy
Cross-region internal proxy Network Load Balancer	Envoy
External passthrough Network Load Balancer	Maglev
Internal passthrough Network Load Balancer	Andromeda

Choose a load balancer

To determine which Cloud Load Balancing product to use, you must first determine what traffic type your load balancers must handle. As a general rule, you'd choose an Application Load Balancer when you need a flexible feature set for your applications with HTTP(S) traffic. And you'd choose a Network Load Balancer when you need TLS offloading at scale or support for UDP, or if you need to expose client IP addresses to your applications.

You can further narrow down your choices depending on your application's requirements: whether your application is external (internet-facing) or internal, whether you need backends deployed globally or regionally, and whether you need Premium or Standard Network Service Tier.

The following diagram shows all of the available deployment modes for Cloud Load Balancing. For more details, see the Choose a load balancer guide.

Summary of types of Google Cloud load balancers

The following table provides details, such as the network service tier on which each load balancer operates, along with its load balancing scheme.

Load balancer	Deployment mode	Traffic type	Network service tier	Load-balancing scheme *
Application Load Balancers	Global external	HTTP or HTTPS	Premium Tier	EXTERNAL_MANAGED
	Regional external	HTTP or HTTPS	Premium or Standard Tier	EXTERNAL_MANAGED
	Classic	HTTP or HTTPS	Global in Premium Tier Regional in Standard Tier	EXTERNAL†
	Regional internal	HTTP or HTTPS	Premium Tier	INTERNAL_MANAGED
	Cross-region internal	HTTP or HTTPS	Premium Tier	INTERNAL_MANAGED
Proxy Network Load Balancers	Global external	TCP with optional SSL offload	Premium Tier	EXTERNAL_MANAGED
	Regional external	TCP	Premium or Standard Tier	EXTERNAL_MANAGED
	Classic	TCP with optional SSL offload	Global in Premium Tier Regional in Standard Tier	EXTERNAL
	Regional internal	TCP without SSL offload	Premium Tier	INTERNAL_MANAGED
	Cross-region internal	TCP without SSL offload	Premium Tier	INTERNAL_MANAGED
Passthrough Network Load Balancers	External Always regional	TCP, UDP, ESP, GRE, ICMP, and ICMPv6	Premium or Standard Tier	EXTERNAL
	Internal Always regional	TCP, UDP, ICMP, ICMPv6, SCTP, ESP, AH, and GRE	Premium Tier	INTERNAL

Interfaces

You can configure and update your load balancers by using the following interfaces:

• The Google Cloud CLI: A command-line tool included in the Google Cloud CLI; the documentation calls on this tool frequently to accomplish tasks. For a complete overview of the tool, see the gcloud CLI guide. You can find commands related to load balancing in the gcloud compute command group.

  You can also get detailed help for any gcloud command by using the --help flag.

  gcloud compute http-health-checks create --help
  
  

• The Google Cloud console: Load-balancing tasks can be accomplished by using the Google Cloud console.

• The REST API: All load-balancing tasks can be accomplished by using the Cloud Load Balancing API. The API reference docs describe the resources and methods available to you.

• Terraform: You can provision, update, and delete the Google Cloud load-balancing infrastructure by using an open source infrastructure-as-code tool such as Terraform.

What's next

• To help you determine which Google Cloud load balancer best meets your needs, see Choose a load balancer.
• To see a comparative overview of the load-balancing features offered by Cloud Load Balancing, see Load balancer feature comparison.