With ETNO’s call for a so called “fair share” of the profits of large content providers, and the recent political interest around this proposal, we, as small non-profit ISPs want to provide our view of the IP interconnection market in order to contribute to the debate.

Who are we ?

The FFDN, or FDN Federation, is a federation of small ISPs which have in common the absence of redistribution of profits, only seek to improve the network and the service provided, and to strictly abide by the rules of net neutrality. In short, ISPs for their users, by their users.

Our federation has a very wide diversity of members: some provide Internet access to only a few municipalities (SamesWireless, Stolon, Tetaneutral, …), others provide Internet access to students (Rezel, Aurore), others are nation-wide ISPs (FDN, Neutrinet, SNN, …) and yet we face very similar situations and problems.

State of interconnections

Ease of interconnection

Most members are not big enough to manage a large number of interconnections, so they instead rely on transit and mutualize resources to manage them. Some members, however, have access to an IXP which dramatically increases the number of interconnections.

The main contributor to interconnection diversity in our members is whether or not they have the ability to be present (ie have a PoP) in a datacenter, thus putting extra strain on small-scale ISPs operating in rural areas, often far from datacenters and without pre-existing optical infrastructure to carry data to them.

Members relying on third party backhaul operators to carry traffic from diverse and remote locations to a shared PoP often benefit from being in a datacenter, but often without the scale to bear the costs alone and therefore mutualise the resources such as colocation space and interconnections. The Gitoyen association for example, is a structure with such endeavours that some of our members are part of, and smaller scale, bilateral arrangements between members can be made.

On IXPs, the difference in policy between content providers, large and small, and large ISPs is stark : few of the large ISPs have an open peering policy, and when they are forced to peer (by law in the case of France), the contractual conditions are much more convoluted than the usual “send an email/request” we see of most content providers. Moreover few large ISPs are present on route servers, while a lot of content providers are. Route servers are, simply put, an interconnexion hub in an IXP, where one connection with the route server interconnects one to all or most of the peers connected to it.

This casualness of peering interactions, with a very light contractual framework, often nothing more than a few email exchanges, and the respective peering policies, is part of what makes the internet robust : it encourages diverse interconnexions, thus strenghening the internet’s resilience.

Transit and peering prices and conditions

Because of our scale and our values, the practice of paid peering is absent in our interconnections, therefore this section will focus on transit and settlement-free peering.

Transit

As aforementionned, most of our members rely on transit for our interconnection needs, often mutualised in some way with other members.

Prices and conditions vary a lot depending on a presence in a datacenter.

Outside of a datacenter, competition is scarce and our members don’t have any kind of bargaining power. When large operators able to deliver transit via a dedicated fiber deign to reply, no further negociation is possible. It is very rare when we manage to have more than 3 offers.

We have seen prices hover around 1€ per mbps at the 95th percentile, but rarely below, often assorted to a commited contract duration of 3 years or more. We also have to mention that for low traffic volume, as is often the case with our members, to even have an offer is a miracle, and for them to be reasonable for our structures tends to be a swiftly squashed wish.

We acknowledge the needs to account for infrastructure costs, but it needs to be said that while such prices might be considered fair between for-profit companies because of the profit the client would make thanks to the service, such reasonning cannot be applied to non-profit organisations, as no such profit is made.

In a datacenter, the situation is much different. There is far more competition, and the conditions are far better with commitments of a single year being more of the norm. We have seen transit prices from around 0.2€ per mbps at the 95th percentile, to 2€, though prices per mbps tend to increase as commited bandwidth decreases. For our members with low traffic volume, there are far more service providers offering reasonable offers.

Peering

Peering without a PoP in a datacenter is almost non-existent. While it is possible to buy remote peering (a port in an IXP, connected to the client network by rented infrastructure) from large infrastructure service providers, the costs of the service are prohibitively expensive.

Inside a datacenter however, peering becomes reasonable. For the lower tiers of FranceIX, the main french IXP, the publicly available price per mbps is around 0,25€ and in Belgium, with BNIX and BelgiumIX, the price per mbps accessible to our members due to our small size goes around 0.1€.

Some of our structures are able, because of their non-commercial goals, to be sponsored for peering ports at IXPs, but this kind of arrangement is not common, thus our structures often have to bear the full costs or mutualise them. Structures such as Gitoyen for example does peering in addition to purchasing transit and other kinds of arrangements between members to access peering are sometimes made.

While FranceIX is expanding into multiple locations, our members feel a lack of diversity in locations with IXPs, thus encouraging a concentration of core ISP network infrastructure in or near Paris and a dependency on third party backhaul network services to deliver the traffic to a relevant datacenter.

We therefore feel that any policy working towards a diversification of IXP locations would have many benefits:

• increasing Internet resilience, as decentralising the network creates more room to accomodate for unavailability of services at a given location;
• enabling more small networks to interconnect via peering thus strengthening overall resilience of the internet for small ISPs and small and medium businesses and service providers alike;
• increasing digital territoral attractivity of now less attractive areas due to lack of connectivity.

In network CDNs

Some content providers offer ISPs to install cache servers on the ISP’s network to decrease overall bandwidth usage and user to service latency.

However, due to our small traffic volume, such cache servers do not provide the bandwidth usage benefits, as it would take more bandwidth to synchronise it to the content provider’s network than it would save in service to user bandwidth.

Also, such offers are often only made if the ISP’s usual bandwith use with the content provider is above a certain threshold none of our members has reported to meet.

Thus none of them has also reported have an on-network CDN server.

Sender pays, “Fair share” or contribution of content providers to telecom infrastructures

Sender pays, or as the large telco infrastructure operators call it, “fair share”, is a scheme in which large content providers, with highly outbound traffic, pay a fee for this traffic to the receiving ISPs.

The proponents of such a scheme argue that large content providers generate enormous profits from an infrastructure paid for by the ISPs, with an ever increasing bandwidth need due to the content pushed by content providers, while ISPs’ margins get thinner because of competition, thus, according to them, stifling their investment efforts in their networks. They also argue that ISPs do not have the bargaining power necessary to push for such a contribution and ask for regulatory action to be taken to enforce it.

We strongly oppose this scheme, as we deem it a threat to net neutrality, based on incorrect premices, and as the consequences on the internet as a whole would be far-reaching and detrimental to its health, at least in europe.

ISPs generate their revenue by providing to their users a service, that of being able to send and receive data from and to the internet. For ISPs to be paid by content providers to provide the service to their own users would be akin to the postal services being paid by the recipient after it was already paid by the sender. Especially as the content provider only sends the content to the users because they requested such content. As large ISPs account for the near entirety of the market, a handful of large ISPs can hold hostage millions of users in the negociations between them and large content providers, and while some users might change ISPs over the degradation of quality of service due to the absence of direct interconnections, most will not given the administrative burden of changing ISPs.

If such a scheme were to apply only between large telcos operating an access network and large content providers, especially if it were based on traffic levels, ISPs would have an economic incentive to have users prefer large content providers instead of smaller or self operated hosting and social network services. It goes without saying that such an incentive is a threat to net neutrality, a principle we hold dear and defend everyday with our actions. Moreover, especially if only large ISPs were to be the recipients of such a contribution, it would disproportionately benefit very large actors, at the detriment of smaller ISPs such as our members. The argument that this contribution would go to invest in network infrastructure is hard to swallow coming from large for-profit companies, with an interest and duty to direct profits to their shareholders, while, as stated by our charter, our members redirect all eventual profits to the improvement of our networks.

On the other hand, if this proposed contribution were to also apply to the benefit of small ISPs, not only would our previous concerns over incentives against net neutrality also apply, but it would also create incentives for large content providers to ignore peering requests until absolutely necessary. Due to our small userbase, we can easily be ignored by content providers, while large ISPs cannot, as previously explained. Moreover, our members mostly rely on volunteers, and operate on a very small scale, thus making the contract management of such a scheme a nigh impossible task. This would heavily distort the playing field in favor of very large ISPs at the detriment of small scale, non-profit ISPs.

Where are we at ?

As previously stated, our members face many challenges in increasing their interconnection fabric, mostly stemming from difficulties in reaching a datacenter, and reaping the benefits of competition in transit prices. These difficulties can often be attributed to our members’ small size, thus not reaching the critical mass necessary to overcome them. Our members have adopted different strategies to address those challenges, but they are not ideal and some members may not be able to apply them and reach the same level of interconnection as would be preferable.

Given this state of affairs, we are worried about the recent push from ETNO for the sender pays scheme. As previously stated, this scheme would only add up to those difficulties and would only be a step in the wrong direction. Our members would vastly prefer a push to build more diverse interconnections and a less centralised internet.