Let's get real: deterministic peering

In conversations with customers, we are often asked about our capability of providing “deterministic peering” to reduce latency and maximize performance. Our answer? We choose not to.

In this short article, in collaboration with our CTO, Rick Mur, we’ll briefly explain how deterministic deterministic peering is – really, why it’s rarely actually implemented, and how a more agnostic approach to your middle-mile can benefit the performance of your connectivity.

First things first–what’s peering?

On the highway that we call “the Internet,” your data travels through different networks. Often, it goes through several networks before arriving at its destination. That’s how the Internet was designed. And the Border Gateway Protocol (BGP), aka the GPS navigation of the internet, was created with this in mind. Each network on the Internet is known as an Autonomous System (AS), and by connecting all of the different networks (ASs) together, you get… that's right: The Internet. Typically, a regional Internet Service Provider (ISP) will connect to so-called Tier-1 and/or Tier-2 service providers. Tier-1 providers are providers that connect other providers to each other.

Back to Peering. Peering is like a shortcut. It happens between two or more different networks that have agreed to exchange their traffic with each other, skipping the middleman. It’s a direct relation between 2 networks (ASs) on the Internet. Based on the BGP routing policies (GPS navigation settings), these direct connections will, in most cases, be used for exchanging traffic between these networks.

Moving on: what’s deterministic peering?

Deterministic peering is a term coined as more enterprises start using the Internet as a transport for their corporate network traffic. Basically, it’s peering taken to an even more formal (corporate) level.

The shortcut is fixed, with clear processes that have been pre-agreed. Unlike regular peering, which maintains a level of flexibility, deterministic peering is supposed to be planned and predictable.

Why would I want deterministic peering?

The Internet is a great medium for corporate network traffic, albeit secured by SD-WAN or SASE overlays; however, more traditional enterprise WAN networks were used to private (MPLS-based) links, with dedicated bandwidth and a fixed latency between sites.

The Internet, by design–the one we just explained, is a non-guaranteed network. 

With its dynamic nature in mind (the Internet never fully converges), it’s not possible to guarantee latency and throughput between IP address X and IP address Y to be the same all the time. Deterministic peering, then, looks to bring back those guarantees.

“Provider X owns their backbone, ensuring latency guarantees and deterministic peering. Can GNX do that, too?”

Can Provider X do it at all? One more time: the Internet is a platform that, by design, is not guaranteed. Latency guarantees and deterministic peering are never a 100% commitment because the links connecting different ISPs will always be shared. 

Even more importantly, there is no easy way for customers to verify that it is actually being implemented in the way it is agreed upon.

And while using a backbone as a middle-mile can help you optimize latencies, the ability to use all possible paths available on the Internet also has benefits. It provides more flexibility to your connectivity, no lock-in, and the opportunity to always take the most optimal path, depending on the situation.

If a guaranteed network is part of the customer requirement, we recommend a middle-mile solution that is not dependent on a single vendor (e.g. Megaport or Teridion) – and we are always excited to work alongside them for our customers.

The GNX Solution

At GNX, we work closely with our vendors to optimize paths to the fullest extent. Our 24/7 NOC analyses paths where a customer is experiencing higher than expected latency (discovered based on customer feedback or our own monitoring). By doing this and by being transparent in what we can and cannot offer, we push vendors and their upstream providers to update BGP routing policies to reduce latency and improve our customers’ network performance. We have several real customer cases where we were able to bring tremendous improvements in difficult areas and, needless to say, reduce costs by choosing a more efficient approach.

At GNX we take care of our customers without the need to ask. Rather than selling the idea of deterministic peering, we are focused on delivering results. That’s our promise!