The India Internet Foundation (IIFON) runs a community internet exchange in Kolkata to help smaller internet service providers save costs on data transit. Internet exchanges, or IXs, which have played a huge role in reducing network costs for ISPs, are a major part of the modern internet’s infrastructure. MediaNama spoke to Anupam Agrawal, a trustee at IIFON, about why the foundation is running a community exchange, and what that entails. What follows is a slightly edited transcript of the interview.
MediaNama: To start off, what exactly does being a community IX entail? IXs already proclaim their neutrality and cost-effectiveness for ISPs.
Anupam Agrawal: Suppose you are a small ISP, a very small one at that. All the telcos’ lines end 100 kilometres from the city. So, if after that somebody is taking internet to villages, it’s the small ISPs. Let’s assume a small ISP is servicing 100 customers, charging Rs 500 per customer. At the end of the month, they’ll only earn Rs 50,000, of which Rs 35,000 goes to the telco, who provides upstream bandwidth. On top of this, they will have to pay costs for the wires and infrastructure in people’s homes, and the salaries of the people collecting the fees.
For a very small ISP, this business becomes a little profitable, where they can take Rs 15,000-20,000 home when they have somewhere around 300 customers. Till 100–150 customers, it’s actually loss-making. Once the ISP has a cluster of 300 customers or so, that’s when the business becomes profitable.
When 300 customers are there, the revenue would be Rs 1,50,000 a month, out of which Rs 90,000 goes to the upstream provider. The role of an IX is, if the ISP peers there, with an incremental cost of just the port fees, they will be able to service 100 more customers. That is the scale which the ISP gets. These ISPs are non-existent in big IXs that are present. I’m not against the big players, they have their own model.
But who is taking care of these really small ISPs who need this exchange when they’re trying to grow? If you are an ISP with 1,500 or 3,000 customers hooking onto a big exchange, you would benefit, no doubt. But if an ISP is really small and growing, and is taking things forward in rural areas, how do we make them profitable?
On your question of community, here the community means that we don’t own the exchange. There’s an IIFON IXP Working Group, and its members decide on the policies to be implemented in the exchange. There are different kinds of policies. On the technical side, peers can exchange traffic bilaterally;
A can agree that I have Netflix and am ready to give it to
B. Whether that happens bilaterally or multilaterally is up to the participants at IIFON. This is a decision by members of the exchange.
MediaNama: So if one ISP has peering with a content provider, other ISPs can draw traffic from there?
Anupam Agrawal: Let’s say ISP
A has Netflix [caching], and ISP
B wants Netflix as well, and ISP
D also want it. In a multilateral scenario, ISP
A is bound to share everything with
D. But in a bilateral situation,
A says it will share Netflix traffic only with one provider. This is a decision of the people who are in the exchange, not of IIFON management.
At one point, we were working on a switch which was nearing its capacity. Considering the traffic at the exchange, we had to upgrade it. Had I not been a community exchange, we would have invested the money to upgrade it at least a year beforehand. IIFON doesn’t have that kind of money because we are a community exchange. So what we said was, one of you give us a switch for the time being, and in the meantime we will look for options to go forward.
So one ISP provided contributed a switch to the exchange at zero cost, because he was benefitting, and others were also benefitting. In the meantime, ISOC had a plan to donate some switches around the world. So we gave them the graphs, the list of members, and they happily shipped a big Arista switch to us, which landed at our doorstep three weeks ago, in the middle of the lockdown.
And now, the entire team is working on reconfiguring the new switch. So this is what happens in a community exchange; the community works together to see what can be done.
The third is in terms of pricing: how do you price port charges, who decides that? It’s the working group that decides. In IIFON IX, we have a person who does the technical configurations, someone with five years of experience that we pay for 50% of his time. Including electricity, the cost of running the IX comes to Rs 70–75,000 a month. This cost has to be split evenly among members, so each will pay around Rs 22,000.
In fact, initially when we started, this person was an employee of one of the ISPs. So he said, this cost is from my side. The other person gave the switch when it was being upgraded. The third person paid the rent for the data centre where we hosted the switch. So we ended up not charging port charges to any one of them for six months, because everybody was contributing. This is what community is all about. It’s unlike a bigger exchange with a management, an SLA… If the working group decides that there needs to be an SLA in our services, there will be one. If they decide what we have is sufficient, it’s their decision.
Paying for upgrades
Recently, the group decided that they want to do something more for the exchange, so they are going to charge one-time joining fees, and this money will be kept as a fund to upgrade switches. So instead of waiting till when the switch is going to tank and then replacing them at that point, we grow this fund and upgrade when needed. But I have not decided it, the members did. I’m a part of every conversation, but it is not my decision that determines what happens tomorrow.
The group also decided that we need an exchange in Siliguri. But one thing IIFON has said clearly is that we don’t want to be in cities like Mumbai; we want to take this to the next layer of cities. If you see our expansion in coming months, you’ll see we are moving to smaller cities, where it is more beneficial to do that. For that we needed a Class A ISP, because of the legal experience one IX recently had that was very unpleasant; we wanted to avoid that.
IXs don’t really require a license — NIXI doesn’t have one. It operates as per an office order by the Department of Telecommunications. If DoT comes, we will show them this office order and say that if you think we need a license, we will not operate.
We already have a mili-jhuli government. If you pull us into a legal confrontation, I’ll simply close it down. But it’s sure that IXs are required in tier 2 cities now. If the big IXs are moving into tier-2 cities, or people like us are helping in creating that kind of ecosystem, that is when things will flow. If instead of doing that, if the ecosystem puts everything in the legal bench, we’ll simply close it down. That has been our internal stand. We don’t have the resources — our resources are the qualified technical people who have taken things forward. They are simply not interested in going to the legal side of things.
MediaNama: What role does the class A ISP you have on board play in the community exchange?
Anupam Agrawal: Alliance Broadband has content available with them that they are willing to share multilaterally as well as bilaterally with some peers. There are two benefits that Alliance gets out of this arrangement. One is: they are using a Netflix cache, and since it is being used by other ISPs in the city as well because of the peering arrangement, their cache will be filled with diverse content, lowering their own cost of filling that cache.
Say there are ten ISPs with users watching Netflix and their cache gets filled up; their cost of filling up that cache goes down for people there. Alliance is not incurring additional cost here except for the peering charges, and their cost of filling the caches through upstream bandwidth is reduced.
In the eyes of Netflix, Alliance becomes a much stronger partner ISP, because they’re contributing essentially 80% of Netflix content in West Bengal. Then Alliance gets more caches, and the caches gets upgraded. This applies to all content providers.
And what’s in it for us is: we have a class A ISP partnered. We have made sure we don’t land up in a scenario where we are providing transit bandwidth, that we have taken care of. So there we are okay. So even if DoT comes to us, we have a rock-solid arrangement on the ground to defend that initially.
The group also decides who the class A ISP is. Before this, we had a different class-B ISP that said they would provide this benefit to our IX, but then backed out. Then Alliance joined, so they said okay, we’ll do it. Alliance prefers bilateral arrangements with ISPs more than multilateral, which is fine, as long as traffic is exchanged.
At IIFON, internet measurement is one of our fortes, and we are working on that. Now we have a good network across Kolkata to see how DNS traffic is actually flowing. Based on that, one can take good policy decisions either at the state or central level.
MediaNama: Do you think the cost of upstream bandwidth that ISPs buy from telcos is fair in India?
Anupam Agrawal: I think that is not ISPs’ challenge at all. In terms of pricing, the ISP market after the arrival of the bigger player [Jio] in 2016 has gone down. Now for a Rs 500 plan, you are expected to give a lot more than earlier. Every time an ISP upgrades, the rate for the plan goes down. So how do you survive in a market where the price of the same product is going down, and you have to give more than earlier?
And that is the push for peering, peering makes this possible. We can’t blame the cost of upstream bandwidth for this debacle, that price has also commensurately reduced. But imagine being in a business where your prices are going down and consumer expectations are increasing. But this is also an industry where it has become optimise costs and bring profits.
The DoT has become much more vigilant. Earlier some ISPs were doing some idhar-udhar with AGR, but that is very tight now and everyone is more accountable, which is a good thing. Now imagine: they have to upgrade equipment, do security best practices, all sorts of user experience management, and then the sales price goes down…
Saying that the upstream price has to go down just to make small ISPs more profitable, that might be unfair.
MediaNama: Small ISPs in rural areas with limited populations can saturate really quickly. And you can’t increase prices, like you mentioned. So how does an ISP survive then?
Anupam Agrawal: Your question should be: who are these ISPs in rural areas? People who were traditionally operating the government’s Common Service Centres (CSCs), they have suddenly become the ISPs. They offer additional services like filling out forms, facilitate bank transactions, and even phone charging stations, since they have constant power supply. All kinds of innovative models are coming there. On the same line you provide electricity, you put in a switchboard for phone charging. If you go into the hinterland, you find these models. They are much more innovative at the ground.
The second group is cable operators. Internet is recent, but TV has been around for a long time in villages. These are the same people supplying cable connections to houses; they have become ISPs. They are now the owners of two different verticals in a business, supplying internet and TV. Even with 300 connections and small help in terms of peering it is profitable.
Where they get stuck, and that is where the community comes in, is that they are not able to configure the network. The ISP is stuck. Other IXs will not handhold these ISPs to configure their IPv4 or BGP. It is in our mandate to do that, even if they are not a member to help them if they reach out to us. There are people in my exchange who are members, but they have not peered with us. They come to meetings, sign up to our mailing lists, but are located far away — they say if you come to where we are, we will join.
MediaNama: On your PeeringDB, only Akamai is listed as a content provider. Why aren’t content providers moving as aggressively to join the community exchange?
Anupam Agrawal: It’s not that we haven’t approached them; we have. I really don’t know why we get denied. But we are open to the idea of having as many content providers as possible. But perhaps it is the policy of Google or Facebook not to peer at community exchanges; I don’t know what stops them from coming to us. But we have reached out to them.
From some places we got a traffic threshold which we would have to hit first. Some people are like, get to 5 gigs of traffic and we’ll peer. Some places clearly say they don’t want to peer at exchanges.
This is also an emerging trend: content providers don’t want to peer with exchanges, they want to be directly at the level of ISPs. When they peer at exchanges, the hardware is generally given by them. The bandwidth is taken care of by them in some places, and in other places the IX does. They essentially find it leveraging to be directly present at the ISP level, where they have all the bandwidth to fill caches directly. And then, eyeballs are one hop closer.
But our goal is not to have content providers on our lap and run with it. We are comfortable with people coming together to help each other. Making Alliance agree to this model and show them the benefit, it took me a year. It was not like they were not agreeing, but it was a sea change for them to start sharing something that was theirs. Imagine sharing something Google has given you exclusively to others. It’s a change in mindset that takes time and effort to hammer in. We are not in a hurry.
MediaNama: How much data traffic do you have per day?
Anupam Agrawal: Six gigs a day.
MediaNama: What policy steps do you think the government should take to encourage the exchange ecosystem?
Anupam Agrawal: From a policy standpoint, ISPs need to be handheld in terms of security. We are in a mess at the ground level in terms of security. There are not enough people on the ground to configure those kinds of things. If we start some kind of vocational courses on the network layer, there is a huge pool required on the ground. If you see the cable operator business, they would get a satellite dish, install it, and run a cable to everybody’s homes. The TV just worked. And these people have become ISPs.
So some kind of policy intervention is necessary, or we will have cracks.
In this kind of market, if there is a clear policy on IXs themselves, and there is some distinction between community and for-profit IXs, that would be good. Both models exist everywhere in the world. In India, let’s assume there are ten exchanges. This is not good. Unless and until peered content is 80% of an ISP’s traffic and core internet bandwidth use is 20%– that is the ratio in the US. Right now in India, peering is at 40% and broadband is 60%.
We don’t have a study to substantiate this figure, but this is what I understand from people on the ground in terms of their costs and traffic patterns. The more ISPs move towards peering, the more the situation will improve. It is very important that there be a clear-cut IX policy, with a distinction between us and for-profit exchanges, as our purpose is different.
We have tied up with a few educational institutes, where have taken on interns to study traffic patterns on an IX. This may not be the mandate of a big exchange, which is absolutely fine. But our mandate is different. The community value has to be recognised in this entire thing. There may be certain parameters you can decide.
An educational policy is also needed where you train people on the network layer. Engineering colleges don’t teach network switches — it’s not like you can specialise in routing. People have to be pulled in the network layer, that’s where there is an absolute shortage of people.
The government also needs to do something on internet measurements. Tomorrow if you ask anyone in the government, how much traffic goes from Tamil Nadu to Gujarat, nobody can answer that today. Unless you have that kind of visibility, you won’t be able to make concrete policy decisions. Maybe the entire thing is getting chockablocked somewhere near Nagpur or Raipur — that is my assessment.
Now if that is really happening, then Raipur should be the hotbed for datacentres. Can you imagine? Now, this is theoretical. But how this topography actually works, more investment into the network layer in terms of doing some research is required. Policy decisions cannot be made based on your MediaNama report, with someone picking it up and saying “we need IXs”. If we put some focus on internet measurement, or ask ISPs for this data, instead of simply asking for 27001 compliance, we can look at things holistically. Maybe the government needs to engage with some technical guys on the table, rather than just the policy people and the lawyers.
MediaNama: The National Internet Exchange of India has been trying to do the exact same thing you have, in terms of improving connectivity in a geographical spread in an area with exchanges in cities like Guwahati and Ahmedabad, but they have some exchanges that hardly see kilobits of traffic on a daily basis. Do you see a gap to fill there, and do you think they will step into the community role?
Anupam Agrawal: There is a huge amount of work that is left. If they enter the field and say you are competing with NIXI and you should stop, I will swiftly move into the research side of it. That’s the reason we have deployed those two interns, to start the work of studying traffic patterns, which is not in NIXI’s mandate.
Had NIXI done what they were supposed to do, considering the revenue model that existed, I would have happily contributed to that rather than starting something new. Tomorrow, now having the experience of doing this, if NIXI wants to collaborate, that is a fairer game for them to play than try to do something they couldn’t for many years. It takes a lot of energy and effort to do community building, which is difficult for a governmental organisation.
We don’t have to do note-sheets to do our work. We can just call a Zoom meeting and prepare the minutes on WhatsApp, and we’re done for the day. Assuming they have to do this kind of thing, they’ll have to make a note-sheet, process the files.
Maybe it is not the job of the government to do the community building. Just blaming NIXI is not good. Having created NIXI, they should have identified partners, people who can do this legwork and reach out to people more easily and support them. Tomorrow if I want to do a seminar on IPv4 in Begusarai, what will be the expense? Rs 40,000? Rs 30,000? Can they support this? Do they have a policy to do this? They need partners to get these things done, rather than trying to do it on their own.