Article Highlights
- Peering has been moving away from public/IXP peering and toward private peering because the cost of private peering has been declining.
- To manage scale and maintain quality, hyperscalers are driven to peering programs where they select specific certified partners.
- Innovation needs to become more widespread when it comes to cross-connect pricing and the automation of ordering, moving, or cancelling cross-connects.
- As the selected peering partner of Vodafone, Inter.link prioritises quality and transparency, clearly defining capacity planning, utilization limits, and performance expectations upfront.
For over a decade, the fundamentals of peering in Europe have remained largely unchanged — even as bandwidth, hardware, and tooling improved dramatically.
However, recent developments from Google, Vodafone, and others have shifted the European peering landscape. This affects enterprises, ISPs, IXPs, and a whole variety of other businesses that require and provide connectivity.
Let’s explore what is changing for peering, and what still must be done for connectivity to truly thrive across Europe in a way that benefits everyone.
The CEO and Co-Founder of Inter.link, Theo Voss, offers his perspective on this topic, and James Bensley (Tech Lead Product) and Gabor Reich (Team Lead Business Development) provide further technical and commercial insight.
Recent Peering Changes in Europe
When considering the Google VPP Program and Vodafone’s selection of Inter.link to enable interconnection for partners and service providers across core EU markets, this change in approach to peering did not come from nowhere.
For starters, 100G ports are much more readily available. Expensive routers used to have only a few 100G ports, and everything else was 10G. In addition, there has been a change in availability in merchant silicon hardware, technology which greatly improves power and space efficiency in racks in data centres, and there has been a huge reduction in price of IP Transit.
Perhaps most important is that the cost of private peering has been declining, and this has meant that peering has moved more towards private peering, and away from public/IXP peering, due to IXP prices decreasing at a lower rate than the price of private peering.
Unsustainable Pricing for Quality Connectivity
This low price of IP Transit has massively reduced the ability of Tier 1 ISPs to provide congestion-free connectivity.
Whether a customer wants the price to be 10 cents or 2 cents, this is a saving the operator needs to generate on the other side.
When everybody paid 10 cents for 100G commitment, companies could afford to have large blackholes and big backbones, however going down to 2 cents or 3 cents per Mbps leads to everybody trying to save as much as possible and this focus on cost-saving can be damaging for quality.
For example, in the past a PoP could be oversubscribed by only 20%, but now it is common to be oversubscribed by 60% because ISPs cannot afford this change in cost.
If the cost equation becomes unbalanced, then marginal profit is no longer made, bringing about bankruptcy or buyouts. This is an obvious development in the peering world, but it is still neglected by many.
Changes in enterprises but also greater availability of 100G ports have led to many more companies using PNIs (private network interconnections), and IXPs have come under pressure because PNIs are an alternative to peering at an IXP.
However, IXPs are expensive if you compare them at a price per Mbps level.
“Innovation is needed amongst the IXPs regarding how they bill for services if public peering is going to survive. We are now seeing the first IXPs switch from flat port rates to usage-based billing, and to flat rate billing but shared across multiple ports and peering LANs.”
James Bensley – Tech Lead Product
The Challenge of Fully Utilizing Flat Rate Ports
Prices at IXPs are particularly expensive with flat rate ports which can almost never be used to 100%.
For the vast majority of organizations requiring connectivity, it can be a real challenge to make the most of a flat rate (non-burstable) port because port utilization can only be approximately 85% or higher if your technology is extremely powerful, making it precise enough to steer the bits, and forecast the bits before they go on the port, to control the spikes of traffic overall.
Usually, smaller customers can only utilize 40-50% of a port because they have a low ability to control the spikes. As soon as they use 60%, they need to buy another port because otherwise this causes a bigger problem for them in the long run.
It is like paying for a car but only being allowed to use the half the seats – the cost per passenger inevitably increases
This puts IXPs under pressure when it comes to price.
Hyperscalers Are Driven to Peering Programs
Following on from congestion and issues with Tier 1 ISPs, IXPs being very expensive, and everybody trying to use PNIs, the hyperscalers state that their enterprise customers care about quality. On top of this, hyperscalers deal with connectivity on a huge scale; for example, this can even get to millions of PNIs for Google.
All of this needs to be managed, everyone is under cost pressure, and cross connect prices have barely dropped and managing huge estates of cross-connects requires considerable effort so, for hyperscalers the question becomes: “How can we bring the cost down and effort down of peering while we increase reliability, transparency, and quality?”
The obvious next choice is that instead of talking to everyone individually, hyperscalers talk to a couple of partners and outsource the effort to them, giving them certified badges.
For maintaining quality connectivity when it comes to port utilization for example, Google will put a system in place that ensures utilization is never beyond a certain level.
The Google VPP Platinum level will make the utilization of your ports with Google transparent. This will show the end customer how much percentage of traffic is being used, and this will only happen if utilization is kept below 75%.
Vodafone is in a similar position. They have tens of thousands of BGP sessions, a small team, and are under a lot of price pressure from customers.
When costs are going down and the volume of traffic (bandwidth demand) is increasing, who should finance the required hardware?
Overall, the imbalance between price expectation and cost along with growth is a challenge. All of this with quality included as a key focus point means that hyperscalers are driven into peering programs like Google VPP or Vodafone’s new initiative with Inter.link.
“At today’s pricing levels, every cent saved per Mbps has a direct impact on congestion, resilience, and network performance. Buyers need to understand what is given up to achieve ultra-low pricing.”
Gabor Reich – Team Lead Business Development
Changes Needed for the Connectivity Industry
With the evolution of technology going on for years and years, right now it seems like we’re at the peak, and this is the tipping point for the network industry.
Connectivity is very cheap, expectations are very high, and people are getting extremely frustrated.
What kind of changes are needed for connectivity to thrive across Europe and beyond in a way that benefits everyone?
Distinguishing between Quality and Pure Price of Connectivity
The industry needs some consolidation amongst IXPs, possibly some consolidation amongst ISPs, and the network industry needs to differentiate between quality and pure price when it comes to options for connectivity.
It’s all well and good for a peering manager to hunt for the lowest possible price for connectivity and be thrilled to receive an offer for 2 cents per Mbps, but if they are the one delivering the services, they need to understand what this means in the long-term, especially if they care about quality.
Innovation for Cross-Connects and IXPs Overall
Innovation is needed amongst the IXPs regarding how they bill for services if public peering is going to survive. We are now seeing the first IXPs switch from flat port rates to usage-based billing, and to flat rate billing but shared across multiple ports and peering LANs.
We have seen some data centre operators operate their own neutral IXPs, or provide discounted cross-connects to IXPs, and different pricing models based on the kinds of cross-connects needed.
If we want private peering not to be replaced with cheap and low quality IP transit, cross-connect pricing innovation needs to become more wide spread. In addition, automation of ordering/moving/cancelling cross-connects is still in its infancy, something which has been long been possible.
How Inter.link Prioritizes Transparency to Ensure Quality
Inter.link customers benefit from commercial models where capacity planning, utilization limits, and performance expectations are clearly defined upfront. These all help reduce paying for bandwidth that looks cheap on paper but creates operational issues later.
Customers and partners who align realistic cost structures, quality requirements, and growth assumptions will avoid constant renegotiations and build more sustainable connectivity strategies.
This is what Vodafone and Google have begun doing. They decided to shift from transactions to partnerships, where they favor a trusted interconnection partner, such as Inter.link, over hundreds of bilateral relationships. This not only reduces operational effort for them but also improves customer experience.
Peering Moving Forward
The shift to peering programs made by Google and Vodafone is the result of changes in the transit and peering landscapes.
Since it is not sustainable for IP Transit to be priced incredibly low whilst still maintaining high quality connectivity, the network industry needs to differentiate between quality and pure price.
The low rate of decline of IXP pricing, coupled with the high rate of decline in of private peering pricing, has lead to massive investment in private peering which has become unmanageable.
In addition, it is beneficial for the entire community in the long-term if there is greater innovation around cross-connect pricing and the automation of ordering, moving, and cancelling of cross-connects.
