Premium vs Standard tier in GCP

Google, in general, is regarded as one of the most densely interconnected networks in the world, thanks to its vast underground, terrestrial and sub-sea cable networks alongside multiple peering point locations with several ISPs in the world. Google Cloud Platform leverages this network to transport its packets between resources in GCP and clients out on the open Internet. More particularly when it comes to traffic sourced from the Internet towards a resource on GCP, there are two tiers -

• Premium Tier
• Standard Tier

To understand the concept of premium vs standard tier networking in GCP, you must first learn the concept of hot-potato routing and cold-potato routing. These two are routing techniques that are antonyms to one another. A typical path for packets between a given source and destination in two different networks involves at least those two networks in theory and several more in realty. In a hot-potato routing scenario, each network between source and destination will try to handover the packet to the next network as soon as possible. It's the exact opposite in case of cold-potato routing technique. The network, in this scenario, holds on to the packet as much as it can with all next-hops within the same network and hands it off only when it doesn't have a next-hop available that's closer to the destination.

With this in mind, let's look at what Premium and Standard tiers are. Under premium tier, all traffic inbound from the Internet to your resources in GCP (be it VMs with external IP addresses or external load balancers) enter Google's network as close to the client as possible. This means, the packets spend a significant amount of time and hops inside Google's own reliable and low latency network. This is possible because of all the hundreds, if not thousands, of points of presence locations around the world where Google directly peers with external networks. There is a very good chance that the ISP of your client is directly connected to Google's network in at least one location. Due to the redundant nature of Google's backbone, it is highly fault tolerant and almost always results in the best latency possible between that client and the GCP resource when it traverses this network vs. public Internet. Outbound traffic from GCP towards the client always exits through a PoP closest to them.

On the other hand, Standard tier is exactly what it sounds like. This traffic follows the "general" routing pattern across the world. Google accepts traffic from your clients in the region where your resources (VMs or load balancers) are. Outbound traffic from your resources to the client rides the Google network only until the immediate PoP between Google and an upstream carrier in the same region as the resource. The packet then traverses through one or more external networks before ultimately making their way to the client. Since this relies on public Internet and routing stability, there is no expectation of reasonable guarantees of low latency and/or reliability.

This makes Standard tier slightly cheaper than Premium tier. The regional IP addresses associated with your resources can be either premium tier or standard tier. However, some regional products (looking at you, Cloud NAT!) doesn't support this network routing distinction. They are always billed as premium tier IPs.

Google has an excellent flow chart that helps you decide what's best for you. Posting it here for your review -