Load balancer feature comparison

Stay organized with collections Save and categorize content based on your preferences.

This page gives you a comparative overview of the load balancing features offered by Cloud Load Balancing. If you haven't already done so, begin by reading the following:

  • To get an overview of the different load balancing solutions that are available in Google Cloud, see Load balancing overview.
  • To determine which Google Cloud load balancer best meets your application's needs, see Choose a load balancer.

In the following tables, a checkmark indicates that a feature is supported. For more information about a feature, click the info link.

Type of load balancer

Feature Internal HTTP(S) External HTTP(S) Internal TCP/UDP External TCP/UDP Network External SSL Proxy and
TCP Proxy		 Internal TCP Proxy
Global Classic Regional
Proxy info
info
info
info
SSL Proxy info

TCP Proxy info
info
Pass-through info info

Protocols from the load balancer to the backends

For links to reference information, see Backend services.

Feature Internal HTTP(S) External HTTP(S) Internal TCP/UDP External TCP/UDP Network External SSL Proxy and
TCP Proxy		 Internal TCP Proxy
Global Classic Regional
One of:
HTTP (HTTP/1.1)
HTTPS (HTTP/1.1)
HTTP/2 (requires TLS), including gRPC
One of:
TCP or UDP
One of:
SSL (TLS) or TCP
GRE, ESP, ICMP, or ICMPv6 info
WebSockets info
info
info
info

Protocols from the clients to the load balancer

For links to reference information, see Forwarding rules.

Feature Internal HTTP(S) External HTTP(S) Internal TCP/UDP External TCP/UDP Network External SSL Proxy and
TCP Proxy		 Internal TCP Proxy
Global Classic Regional
One of:
HTTP/1.1, HTTP/2, or HTTPS
(no QUIC support)
(includes QUIC)
(includes QUIC)
(no QUIC support)
HTTP/3 (based on IETF QUIC)
One of:
TCP or UDP
SSL or TCP
TCP only
GRE, ESP, ICMP, or ICMPv6 info
WebSockets info info info info

Backends

Feature Internal HTTP(S) External HTTP(S) Internal TCP/UDP External TCP/UDP Network External SSL Proxy and
TCP Proxy		 Internal TCP Proxy
Global Classic Regional
Backends must be in one region
(Standard Tier)
(Standard Tier)
Backends can be in multiple regions
(Premium Tier)
(Premium Tier)
Load balancer can have multiple backend services and a URL map
Virtual machine backends on Compute Engine
Zonal NEGs
Using GCE_VM_IP_PORT type endpoints with GKE:
Using GCE_VM_IP_PORT type endpoints with GKE:
Using GCE_VM_IP_PORT type endpoints with GKE:
Using GCE_VM_IP type endpoints with GKE: 		Use standalone zonal NEGs
GCE_VM_IP_PORT type endpoints
Self-managed Kubernetes and GKE
Supported only with the GKE Gateway controller (Preview)
External endpoints in internet NEGs info
(Premium Tier)
Private external endpoints in hybrid NEG backends info info info info info info
Cloud Storage in backend buckets info info
Serverless backends info
  • Cloud Run
(Preview)		 info
  • Cloud Run
  • App Engine
  • Cloud Functions
 info
  • Cloud Run
  • App Engine
  • Cloud Functions
 info
  • Cloud Run
(Preview)
Private Service Connect NEGs
Access Google API
Access published services
Publish hybrid services

Health checks

For links to reference information, see Health checks.

Feature Internal HTTP(S) External HTTP(S) Internal TCP/UDP External TCP/UDP Network External SSL Proxy and
TCP Proxy		 Internal TCP Proxy
Global Classic Regional
gRPC health checks 3 3 1 3
HTTP health checks 2 2 2 1
HTTPS health checks 2 2 2 1
HTTP/2 health checks 1
SSL health checks 1
TCP health checks 1
Configurable health checks:
  • Port
  • Check intervals
  • Timeouts
  • Healthy and unhealthy thresholds
Configurable request path (HTTP, HTTPS, HTTP/2) 1
Configurable request string or path (TCP or SSL) 1
Configurable expected response string 1

1 This table documents health checks supported by backend service-based network load balancers. Target pool-based network load balancers only support legacy HTTP health checks.

2 The regional external HTTP(S) load balancer doesn't support legacy health checks. The global external HTTP(S) load balancer and the global external HTTP(S) load balancer (classic) support legacy health checks only if both of the following are true:

  • The backends are instance groups.
  • The backend VMs serve traffic that uses the HTTP or HTTPS protocol.

3 If you are allowlisted to use Envoy distributed health checks, then regional external HTTP(S) load balancers, internal HTTP(S) load balancers, and internal regional TCP proxy load balancers that use hybrid NEG backends do not support gRPC health checks. For more information, see the Hybrid NEGs overview.

IP addresses

For links to reference information, see Addresses.

Feature Internal HTTP(S) External HTTP(S) Internal TCP/UDP External TCP/UDP Network External SSL Proxy and
TCP Proxy		 Internal TCP Proxy
Global Classic Regional
Internal IP address, accessible in your Virtual Private Cloud (VPC) network
External IP address (global anycast)
(Premium Tier)
(Premium Tier)
External IP address (regional)
(Standard Tier)
(Standard Tier)
External IP address from Bring your own IP (BYOIP)
Multiple forwarding rules with the same IP address, each having a unique protocol and port combination
Internet accessible
(including by clients that are in Google Cloud and have internet access)
Privately accessible
Client source IP address preservation X-Forwarded-For header X-Forwarded-For header X-Forwarded-For header X-Forwarded-For header In TCP Proxy header In TCP Proxy header
IPv6 support
IPv6 termination
IPv6 termination
(Premium Tier)
info
(Preview)
info
(Premium Tier)
IPv6 termination
(Premium Tier)

Network topologies

Feature Internal HTTP(S) External HTTP(S) Internal TCP/UDP External TCP/UDP Network External SSL Proxy and
TCP Proxy		 Internal TCP Proxy
Global Classic Regional
Relationships between VPC networks and load balancer backends
Backends must be in the same VPC network
Backends can be located in multiple VPC networks in the same project

(the networks do not have to be connected)
Backends can use a Shared VPC network
Load balancer frontend (URL map) can reference backend services created in different service projects in a Shared VPC network
(cross-project service referencing)
info
info
Client access to load balancers
Google Cloud or on-premises clients must access the load balancer privately by being either in the same VPC network, in a peered VPC network, or in another network connected using Cloud VPN tunnels or Cloud Interconnect attachments (VLANs)
With hybrid NEGs
With hybrid NEGs
With hybrid NEGs
Google Cloud client VMs require external IP addresses or a NAT solution like Cloud NAT to access the load balancer
On-premises client VMs require internet access to access the load balancer
Internet access is not required with hybrid NEGs
Internet access is not required with hybrid NEGs
Internet access is not required with hybrid NEGs
Google Cloud client VMs can be located in any region
If global access is enabled		 (Premium Tier)
If global access is enabled
If global access is enabled
Google Cloud client VMs can be located in any project
(subject to other requirements in this table)

Failover

Feature Internal HTTP(S) External HTTP(S) Internal TCP/UDP External TCP/UDP Network External SSL Proxy and
TCP Proxy		 Internal TCP Proxy
Global Classic Regional
Automatic failover to healthy backends within same region
Automatic failover to healthy backends in other regions
(Premium Tier)
(Premium Tier)
Behavior when all backends are unhealthy Returns HTTP 503 Returns HTTP 503 Returns HTTP 502 Returns HTTP 503 info
(configurable)		 info
(configurable1)		 Traffic dropped Traffic dropped
Configurable standby backends info
(configurable)		 info
(configurable2)
Connection draining on failover and failback info
(configurable)		 info
(configurable3)

This table documents failover as supported by backend service-based network load balancers.

1 When all the backends of a target pool-based network load balancer are unhealthy, traffic is distributed among all backends.
2 Target pool-based network load balancers use backup pools to support failover.
3 Target pool-based network load balancers do not support configuration of connection draining on failover/failback.

Logging and monitoring

Feature Internal HTTP(S) External HTTP(S) Internal TCP/UDP External TCP/UDP Network External SSL Proxy and
TCP Proxy		 Internal TCP Proxy
Global Classic Regional
Byte count metrics info info info info info info
Packet count metrics info info info info
Round trip time or latency metrics info info info info info info
Connection count metrics info info
Connection attribute logs info info
HTTP request count metrics info info
HTTP request and response attribute logs info info

Session affinity

For detailed information, see Session affinity.

For links to reference information, see Backend services.

Feature Internal HTTP(S) External HTTP(S) Internal TCP/UDP External TCP/UDP Network External SSL Proxy and
TCP Proxy		 Internal TCP Proxy
Global Classic Regional
Headers
HTTP cookie
Generated cookie
Client IP, no destination (1-tuple)
CLIENT_IP_NO_DESTINATION 		info (Preview)
Client IP, Destination IP (2-tuple)
CLIENT_IP		 1 1
Client IP, Destination IP, Protocol (3-tuple)
CLIENT_IP_PROTO		 1 1
Client IP, Client Port, Destination IP, Destination Port, Protocol (5-tuple)
CLIENT_IP_PORT_PROTO		 1,2 1,2
None (5-tuple)
NONE		 3 3

1 Setting session affinity is only meaningful if the protocol uses sessions. For example, TCP.

2 If the protocol does not have a concept of ports or if the packet does not carry port information (subsequent UDP fragments, for example), then a 3-tuple hash of the Client IP, Destination IP, and protocol is used instead.

3 If the protocol has a concept of ports and the packet carries port information, then None is a 5-tuple hash. If the protocol does not have a concept of ports or if the packet does not carry port information (for example, subsequent UDP fragments), then None is a 3-tuple hash of the Client IP, Destination IP, and protocol.

Load balancing methods

For detailed information, see the Backend services overview.

For links to reference information, see Backend services.

Feature Internal HTTP(S) External HTTP(S) Internal TCP/UDP External TCP/UDP Network External SSL Proxy and
TCP Proxy		 Internal TCP Proxy
Global Classic Regional
Balancing mode: connection
Balancing mode: rate (requests per second) 2 2
Balancing mode: backend utilization
(instance group backends only) 		2 2
Configurable maximum capacity per backend instance group or NEG 2 2
Circuit breaking 2
Percent of traffic/weight-based
Prefers region closest to client on the internet

When the closest region is at capacity or isn't healthy, prefers next closest region
(Premium Tier)
(Premium Tier)
Within zone/region load balancing policy
Load balancing locality policy 2
Round robin in a zone and load balancing locality policy
Round robin in a zone
Round robin in a zone and load balancing locality policy
Hash-based distribution among all backends in the active pool when failover is configured, or among all backends in the region
Hash-based distribution among all backends in the active pool when failover is configured, or among all backends in the region 1
Round robin in a zone
Round robin in a zone and load balancing locality policy

1 This table documents load balancing methods supported by backend service-based network load balancers. Target pool-based network load balancers also perform hash-based distribution among all instances in the target pool or backup pool.

2 This feature is not supported with load balancers using serverless NEG backends.

Routing and traffic management

Feature Internal HTTP(S) External HTTP(S) Internal TCP/UDP External TCP/UDP Network External SSL Proxy and
TCP Proxy		 Internal TCP Proxy
Global Classic Regional
HTTP/Layer 7 request routing Suffix, prefix, and match on the following:
 Suffix, prefix, and match on the following:
 Suffix, prefix, and match on the following:
 Suffix, prefix, and match on the following:
Fault injection info
Configurable timeouts info1 info info info info info
Retries info1 info info info
Redirects info info info info
URL rewrites info info info info
Request and response header transformations
(configured on the URL map)		 info info info
Traffic splitting info
Traffic mirroring info1 info info
Outlier detection info info info
Retry failed requests info info info
1 This feature is not supported with load balancers using serverless NEG backends.

For internal HTTP(S) load balancers, see the following:

For the global external HTTP(S) load balancer, see the following:

For the global external HTTP(S) load balancer (classic), see the following:

For the regional external HTTP(S) load balancer, see the following:

For traffic management features available with Traffic Director, see Traffic Director features: Routing and traffic management.

Autoscaling and autohealing

Feature Internal HTTP(S) External HTTP(S) Internal TCP/UDP External TCP/UDP Network External SSL Proxy and
TCP Proxy		 Internal TCP Proxy
Global Classic Regional
Managed instance group autoscaling based on load balancer serving capacity
Autohealing (native to managed instance groups and GKE)
Connection draining 1
1 This table documents autoscaling and autohealing features supported by backend service-based network load balancers. Target pool-based network load balancers do not support connection draining.

Security

Feature Internal HTTP(S) External HTTP(S) Internal TCP/UDP External TCP/UDP Network External SSL Proxy and
TCP Proxy		 Internal TCP Proxy
Global Classic Regional
Managed certificates info info info
(SSL proxy only)
CORS info
Identity-Aware Proxy (IAP) info
Google Cloud Armor info info info (Preview) info
SSL offload
(SSL proxy only)
SSL policies
(TLS version and cipher suites)		 info info info info info
(SSL proxy only)

Special features

Feature Internal HTTP(S) External HTTP(S) Internal TCP/UDP External TCP/UDP Network External SSL Proxy and
TCP Proxy		 Internal TCP Proxy
Global Classic Regional
Cloud CDN info info
(Premium Tier)
External endpoints in internet NEGs as external backends for Cloud CDN info
(Premium Tier)
Media CDN info
(Premium Tier)
Internal DNS names info
Load balancer as next hop info
Specify network interface of a backend VM
(Multi-NIC load balancing) 		The backend VM's nic0 must be in the same network and region used by the forwarding rule. The load balancer only sends traffic to the first network interface (nic0), whichever VPC network that nic0 is in. info The load balancer only sends traffic to the first network interface (nic0) of the backend VM. The backend VM's nic0 must be in the same network and region as the forwarding rule. The load balancer only sends traffic to the first network interface (nic0), whichever VPC network that nic0 is in. The backend VM's nic0 must be in the same network and region used by the forwarding rule.
Custom request and response headers
(configured on the backend service)		 info info
Automatic Service Directory registration (Preview) info info
Connection tracking policy 1
Source IP-based traffic steering (Preview) 1 info
Weighted load balancing 1 info

1 This table documents features supported by backend service-based network load balancers. Target pool-based network load balancers do not support these features.

What's next

For detailed information on each load balancer, see the following: