> ## Documentation Index
> Fetch the complete documentation index at: https://help.nops.io/llms.txt
> Use this file to discover all available pages before exploring further.

# Use Standard Network Tier for Latency-Tolerant Internet Egress

> Send latency-tolerant internet traffic from VMs and regional load balancers over the Standard network tier instead of Premium.

Google Cloud offers two network tiers for internet traffic. Premium, the default, carries traffic over Google's own network close to the user. Standard hands it to the public internet near the source region and costs less per GiB. nOps finds projects sending Premium-tier internet egress from VM external IPs and regional load balancers in North America and Europe, and estimates the saving from using Standard tier for that traffic.

| | |
| - | - |
| **Cloud** | Google Cloud |
| **Category** | Networking |
| **Service** | Network Service Tiers |
| **Action** | Modify |
| **Effort** | Medium |
| **Restart needed** | No |
| **Can be rolled back** | Yes |
| **Analysis window** | Last 7 days |
| **Source** | nOps |

## Why it matters

Internet egress is billed per GiB, so it grows with traffic. Batch downloads, backups and other traffic that is not latency-sensitive do not need Premium routing.

## How nOps finds it

nOps reads Premium-tier internet egress charges from VM external IPs and regional forwarding rules, grouped by project and source region. A group is flagged when it has been billed for at least 7 days and in the last 2 days, has no global load balancer egress (global load balancers cannot use Standard tier), and its source region is in North America or Europe. The saving is the egress cost over the last 7 days, scaled to a month, times 1 minus 0.065 / 0.08 (Standard against Premium list price per GiB at the highest-volume tiers, about 19%). It must be at least \$5 a month. Projects whose spend on the service is at least 90% paid by credits or promotions are left out.

## Before you start

Confirm the traffic can tolerate higher latency and jitter for users far from the source region.

## How to implement

1. Identify which VMs and regional load balancers send the traffic, and check that it is not latency-sensitive.
2. **VMs:** replace the external IP's access config with a Standard-tier one: `gcloud compute instances delete-access-config VM_NAME --access-config-name='External NAT' --zone=ZONE`, then `gcloud compute instances add-access-config VM_NAME --network-tier=STANDARD --zone=ZONE`.
3. **Regional load balancers:** create a new forwarding rule (and address) with `--network-tier=STANDARD`, move DNS to it, then delete the old one.
4. Optionally set the project default for new resources: `gcloud compute project-info update --default-network-tier=STANDARD`.

## Risks and considerations

* Standard tier traffic uses the public internet, so latency and jitter are higher for distant users.
* Changing the tier usually changes the external IP address. A Premium static address cannot be converted, so reserve a Standard one and update DNS and allow lists.
* Global load balancers and Cloud CDN only work on Premium.
* To roll back, switch the resource back to Premium tier.

## FAQs

> [accordion] Why does nOps use the highest-volume price tiers?
>
> To avoid overstating the saving. At lower volumes the gap between the two tiers is larger, so the real saving can be higher.

## Related

* [Release Idle GCP Static IP Address](https://help.nops.io/inform/supported-recommendations/release-idle-gcp-static-ip-address)
* [Supported recommendations](https://help.nops.io/inform/supported-recommendations.md)


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.