External TCP/UDP Network Load Balancing overview

Google Cloud external TCP/UDP Network Load Balancing (after this referred to as Network Load Balancing) is a regional, non-proxied load balancer.

Network Load Balancing distributes traffic among virtual machine (VM) instances in the same region in a Virtual Private Cloud (VPC) network. A network load balancer directs TCP or UDP traffic across regional backends.

The network load balancer supports any and all ports. You can use Network Load Balancing to load balance TCP and UDP traffic. Because the load balancer is a pass-through load balancer, your backends terminate the load-balanced TCP connection or UDP packets themselves. For example, you might run an HTTPS web server on your backends and use a Network Load Balancing to route requests to it, terminating TLS on your backends themselves.

For more information about supported ports, see the summary of Google Cloud load balancers.

Network Load Balancing has the following characteristics:

  • Network Load Balancing is a managed service.
  • Network Load Balancing is implemented by using Andromeda virtual networking and Google Maglev.
  • The network load balancers are not proxies.
  • 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.
  • The load balancer preserves the source IP addresses of packets.
  • The destination IP address for packets is the regional external IP address associated with the load balancer's forwarding rule.

Consequently:

  • Instances that participate as backend VMs for network load balancers must be running the appropriate Linux guest environment, Windows guest environment, or other processes that provide equivalent functionality.

    The guest OS environment (or an equivalent process) is responsible for configuring local routes on each backend VM. These routes allow the VM to accept packets that have a destination that matches the IP address of the load balancer's forwarding rule.

  • On the backend instances that accept load-balanced traffic, you must configure the software to bind to the IP address associated with the load balancer's forwarding rule (or to any IP address, 0.0.0.0/0).

The following diagram shows users in California, New York, and Singapore. They're all connecting into their backend resources, which are myapp, test, and travel. When a user in Singapore connects into the U.S. West backend, the traffic ingresses closest to Singapore because the range is anycasted. From there, the traffic is routed to the regional backend.

Three regional backends and three forwarding rules (click to enlarge)
Network Load Balancing example (click to enlarge)

For information about how the Google Cloud load balancers differ from each other, see the following documents:

Protocols, scheme, and scope

Each network load balancer supports either TCP or UDP traffic (not both).

A network load balancer balances traffic originating from the internet.

The scope of a network load balancer is regional, not global. This means that a network load balancer cannot span multiple regions. Within a single region, the load balancer services all zones.

Use cases

Use Network Load Balancing in the following circumstances:

  • You need to load balance UDP traffic, or you need to load balance a TCP port that isn't supported by other load balancers.
  • It is acceptable to have SSL traffic decrypted by your backends instead of by the load balancer. The network load balancer cannot perform this task. When the backends decrypt SSL traffic, there is a greater CPU burden on the VMs.
  • Self-managing the load balancer's SSL certificates is acceptable to you. Google-managed SSL certificates are only available for HTTP(S) Load Balancing and SSL Proxy Load Balancing.
  • You need to forward the original packets unproxied.
  • You have an existing setup that uses a pass-through load balancer, and you want to migrate it without changes.

Architecture

The load balancer is made up of several configuration components. A single load balancer can have the following:

A network load balancer always has one target pool. Multiple forwarding rules can reference the target pool.

The target pool is a backend for the load balancer. It specifies the backend instances among which traffic is load balanced. Each forwarding rule is a frontend for the load balancer. Keep in mind that there are limits to the number of forwarding rules and target pools per project.

The network load balancers balance the load on your systems based on incoming IP protocol data, such as address, port, and protocol type.

The network load balancer is a pass-through load balancer, so your backends receive the original client request. The network load balancer doesn't do any Transport Layer Security (TLS) offloading or proxying. Traffic is directly routed to your VMs.

When you create a forwarding rule for the load balancer, you receive an ephemeral virtual IP address (VIP) or reserve a VIP that originates from a regional network block.

You then associate that forwarding rule with your target pools. The VIP is anycasted from Google's global points of presence, but the backends for a network load balancer are regional. The load balancer cannot have backends that span multiple regions.

You can use Google Cloud firewalls to control or filter access to the backend VMs.

The network load balancer examines the source and destination ports, IP address, and protocol to determine how to forward packets. For TCP traffic, you can modify the forwarding behavior of the load balancer by configuring session affinity.

Load distribution algorithm

By default, to distribute traffic to instances, the session affinity value is set to NONE. Cloud Load Balancing picks an instance based on a hash of the source IP and port, destination IP and port, and protocol. This means that incoming TCP connections are spread across instances, and each new connection may go to a different instance. All packets for a connection are directed to the same instance until the connection is closed. Established connections are not taken into account in the load balancing process.

Regardless of the session affinity setting, all packets for a connection are directed to the chosen instance until the connection is closed. An existing connection has no impact on load balancing decisions for new incoming connections. This can result in an imbalance among backends if long-lived TCP connections are in use.

You can choose a different session affinity setting if you need multiple connections from a client to go to the same instance.

Target pools

A target pool resource defines a group of instances that should receive incoming traffic from forwarding rules. When a forwarding rule directs traffic to a target pool, Cloud Load Balancing picks an instance from these target pools based on a hash of the source IP and port and the destination IP and port. For more information about how traffic is distributed to instances, see the Load distribution algorithm section in this topic.

Target pools can only be used with forwarding rules that handle TCP and UDP traffic. For all other protocols, you must create a target instance. You must create a target pool before you can use it with a forwarding rule. Each project can have up to 50 target pools.

If you intend for your target pool to contain a single VM instance, you should consider using the protocol forwarding feature instead.

Network Load Balancing supports Cloud Load Balancing Autoscaler, which allows users to perform autoscaling on the instance groups in a target pool based on backend utilization. For more information, see Scaling based on CPU utilization.

Forwarding rules

Forwarding rules work in conjunction with target pools to support load balancing. To use load balancing, you must create a forwarding rule that directs traffic to specific target pools. It is not possible to load balance your traffic without a forwarding rule.

Each forwarding rule matches a particular IP address, protocol, and optionally, port range to a single target pool. When traffic is sent to an external IP address that is served by a forwarding rule, the forwarding rule directs that traffic to the corresponding target pool.

If you are load balancing UDP packets that are likely to be fragmented before arriving at your Google Cloud VPC network, see Load balancing and fragmented UDP packets.

Multiple forwarding rules

You can configure multiple regional external forwarding rules for the same external TCP/UDP network load balancer. Optionally, each forwarding rule can have a different regional external IP address, or multiple forwarding rules can have the same regional external IP address.

Configuring multiple regional external forwarding rules can be useful for these use cases:

  • You need to configure more than one external IP address for the same target pool.
  • You need to configure different port ranges or different protocols by using the same external IP address for the same target pool.

When using multiple forwarding rules, make sure that you configure the software running on your backend VMs so that it binds to all necessary IP addresses. This is required because the destination IP address for packets delivered through the load balancer is the regional external IP address associated with the respective regional external forwarding rule.

Health checks

Health checks ensure that Compute Engine forwards new connections only to instances that are up and ready to receive them. Compute Engine sends health check requests to each instance at the specified frequency. After an instance exceeds its allowed number of health check failures, it is no longer considered an eligible instance for receiving new traffic. Existing connections are not actively terminated, which allows instances to shut down gracefully and close TCP connections.

The health checker continues to query unhealthy instances, and returns an instance to the pool when the specified number of successful checks occur. If all instances are marked as UNHEALTHY, the load balancer directs new traffic to all existing instances.

Network Load Balancing relies on legacy HTTP health checks to determine instance health. Even if your service does not use HTTP, you must run a basic web server on each instance that the health check system can query.

Legacy HTTPS health checks aren't supported for network load balancers and cannot be used with most other types of load balancers.

Return path

Google Cloud uses special routes not defined in your VPC network for health checks. For more information, see Load balancer return paths.

Firewall rules

Health checks for network load balancers are sent from these IP ranges. You'll need to create ingress allow firewall rules that permit traffic from those ranges.

Network Load Balancing is a pass-through load balancer, which means that your firewall rules must allow traffic from the client source IP addresses. If your service is open to the internet, it is easiest to allow traffic from all IP ranges. If you want to restrict access so that only certain source IP addresses are allowed, you can set up firewall rules to enforce that restriction, but you must allow access from the health check IP ranges.

For a firewall rule example and a configuration example, see Rules for Network Load Balancing.

Session affinity

Network Load Balancing doesn't use backend services session affinity. Instead, network load balancers use target pools for session affinity.

For more information, see the sessionAffinity parameter in Using target pools.

Load balancing and fragmented UDP packets

If you are load balancing UDP packets, be aware of the following:

  1. Unfragmented packets are handled normally in all configurations.
  2. UDP packets may become fragmented before reaching Google Cloud. Intervening networks may wait for all fragments to arrive before forwarding them, causing delay, or may drop fragments. Google Cloud does not wait for all fragments; it forwards each fragment as soon as it arrives.
  3. Because subsequent UDP fragments do not contain the destination port, problems can occur in these situations:

    • If the target pools session affinity is set to NONE (5-tuple affinity), the subsequent fragments may be dropped because the load balancer cannot calculate the 5-tuple hash.
    • If there is more than one UDP forwarding rule for the same load-balanced IP address, subsequent fragments may arrive at the wrong forwarding rule.

If you expect fragmented UDP packets, do the following:

  • Set session affinity to CLIENT_IP_PROTO or CLIENT_IP. Do not use NONE (5-tuple hashing). Because CLIENT_IP_PROTO and CLIENT_IP do not use the destination port for hashing, they can calculate the same hash for subsequent fragments as for the first fragment.
  • Use only one UDP forwarding rule per load-balanced IP address. This ensures that all fragments arrive at the same forwarding rule.

With these settings, UDP fragments from the same packet are forwarded to the same instance for reassembly.

What's next