IEIC Virtual Summit Series #1 “How Resilient is the Internet Infrastructure?” Q&A

June 2nd, 2pm-3pm EDT


  • Dr. Vint Cerf, Chief Internet Evangelist, Google [Keynote]         VGC
  • Andrew Dugan, CTO CenturyLink                                               AD
  • Jason Black, Head of Global Network Infrastructure UBER        JB
  • Staffan Göjeryd, CEO Telia Carrier                                              SG
  • Steve Alexander, CTO Ciena                                                       SBA
  • Ivo Ivanov, CEO DE-CIX International                                         II
  • Jeff Lemmer, CIO Ford Motor Company                                     JL
  • Vinay Kanitkar, CTO Akamai                                                       VK


  1. What potential architecture changes can be expected in another 10 years (e.g. impact of 5G, impact of rogue state controls (NK, Russia sponsored or states which exert non-traditional controls), and impact of demands of new applications e.g. medicine, connected cars, autonomous cars)?

VK: The two big things are–Edge Computing and Decentralization. Edge computing enables us to put compute deeper in the network, closer to the devices that are generating the data and need decisions to be made on it, whether those are simple decisions about aggregation or what data to forward where, or whether it is Machine Learning–about answers, inference you want from some engines in the network. Looking at decentralization, once you get connected to the Internet, you can reach the rest of the Internet through multiple paths, multiple access technologies and essentially be able to create a failover mechanism that truly scales with the needs of the Internet.

SBA: When I look at at the changes coming, I put it into three buckets–(1) the network inherently has to get faster 400Gbps, 800Gbps, 1Tbps (2) Mean Time To Cloud, usually few milliseconds; the edge of the cloud, the network is going to bring it closer. The cloud edge where you will come out of the typical carrier infrastructure and go into dedicated cloud providers — for example a medical cloud, education cloud ; dedicated cloud environments (3) Smarter — automated and seamless so that we can use it everywhere; that is a key piece to resilience. Multiple levels of connectivity is going to be important for us, in order to stay connected all the time.

II: Two main trends–one is we will see creation of more and more specific different interconnection environments; not what we have known for the past 25 years; we will see more segment specific interconnection communities. We are headed towards interconnection platforms for closed user group based on the needs of a single organisation or industry sector; federation for the needs of a single car manufacturer, a big healthcare company or a subset of them having their own security and compliance needs. The second big trend–we call it the Edge Triangle of Interconnection–a mix of Artificial Intelligence, IoT based on 5G interconnection modular approach, with many interconnection hubs close to a cell tower or highway crossroad.

JB: Capacity is the “third” element that we see as a need. If we see the autonomous driving vehicles, for us [Uber] we are looking at the big boxes companies to provide us with parking lot space to off-load all that data and speed is great, capacity is great to take all the data and put it in the cloud, or put it to one of our on-prem data center.

  1. To Vint’s point about home networks being stretched, I haven’t seen a single speaker’s video actually come through smoothly, almost all of them have been fully frozen. Fascinating observations around current usage of the internet since the start of Covid-19.  I would love to hear more regarding current resiliency of the internet as well as the way to move forward and create MORE resiliency. Which could be caused by many places in the network, the WebEx servers, or even on the edge where we have more and more applications running?

VGC: A lot of the ‘choking’ shows up at the edge. There is a peculiar phenomenon ‘bufferbloat’; it teaches you that “more is less”. If you have too much buffest space, you end up buffering too much, and end up with broken video and potentially broken audio as well. The edge access is often a determinant of the quality of the video and audio that you get. So it is important to improve the edge capacity. It is also important to think about what is going on inside a residence or an office building. A lot of people keep their routers for decades; they may replace their mobile phones every year or two, they may replace their laptops with similar frequency but home routers and other wifi equipment almost never gets updated unless it becomes hopeless. This is like having ‘building codes’ for the Internet. We should have building codes for people’s home networks.

JL: That is true; our employees are going to expect that level of service; just as we spend time supporting our enterprises in terms of diagnosing issues, we are going to need greater visibility into how the networks are able to diagnose issues and work with the carriers to resolve them. So the two lanes in the operational components have got to get way better as well.

VGC: Shifting gears to Security; in addition to performance we have serious Security problems at the edges of the net–whether that is a laptop with an OS that hasn’t been updated or mobile with an old OS or a router etc. We also have to think about how do we take care of small, medium size enterprises and residential customers in terms of protecting their safety and security in the network especially if they are working from home then the enterprises need to be concerned about safety and security of residential use. We are wide open in many places. Most users do not have the expertise to respond to that, which makes me wonder if we need cyber fire department or they boy scouts or the girl scouts who would come in help you secure your home network.

  1. What needs to happen to allow consumers to have multiple connectivity options, i.e. more competition to give everyone access at a reasonable cost? What are the implications of that? And how do we do that?

    The most obvious way to give someone real multiple options is to make sure that they are all actually having their own routable IP address space. That would allow them to be connected to more than one network at the same time and be transparent to which direction the traffic flowed in. That would be a configuration nightmare that none of us would want to figure out how to manage. Either that or network mapping tricks can be employed to make a private network look the same from a user’s point of view, no matter which way the traffic goes out on different IP address spaces.

AD: Being connected to multiple carriers can get challenging in the Internet with routing of traffic across multiple carriers. Another angle on this question is purely competition–how do you get multiple consumer choices to choose who your provider is. In some parts of the country, there is choice, in other parts of the country there isn’t. We have an issue in the United States, where rural parts of the country or sometimes even cities haven’t made a lot of investment in networks. That is a result of most network providers are for-profit businesses and have to have an economic incentive to invest. The government is helping with that, they do have funds that have been established. The one that is ending next year or so is called the — ConnectAmerica fund, where the government subsidizes builds in underserved areas. The next generation of that is called the Rural Digital Opportunity fund. It is something similar where companies can bid on the subsidies for building out to the underserved or uncompetitive areas. Hopefully those government subsidized programs will improve the capacity in those underserved areas.

  1. Would today’s global Internet that we rely on for more and more critical applications be impacted if a main interconnection peering point was lost due to a natural or man-made disaster? Are there enough interconnection peering points globally?

    VGC: The answer is ‘no’, there aren’t enough interconnection hubs around the world where peering takes place. We don’t have as much connectivity and interconnection as we need.

II: Our mandate as an industry, as a community is to increase the interconnection hubs, it is not about exchange points only; we see our mandate to localize interconnection as much as possible. This is what we believe should happen even more and more in future, in different geographies that do not enjoy today this type of infrastructure locally and they are disadvantaged. We hope this changes in the coming years and we do it by establishing more and more digital infrastructure in more regions and more geographies. We also believe that the congestion will be avoided by creating huge aggregation levels. We will need modular approaches, edge pre-aggregation nodes which will handle different types of interconnection needs, IoT communication and also 5G traffic that we will see in the next few years.

  1. We all expect 5G to transform the way we use the internet…but do we think that it could stress internet infrastructure. If so, how could we account for that potential increase in demand?

VGC: I am not so persuaded about this. I do think current 5G standards for fine-grained control will lead to efforts by wireless ISPs to introduce differential charging in an attempt to re-capture value-added revenues that mostly flow to content providers at present. Not necessarily a good thing for consumers.

VK: Traditionally, demand has been driven by number of users x use-cases x connection bandwidth distribution. In my opinion, 5G doesn’t change that fundamentally from a human usage perspective. The Internet infrastructure last mile and core will scale up over time to support the transition to 5G. The big trend that’s different now (with or without 5G) is the internet of things where we’re seeing a rapid increase in the number of devices accessing the Internet. All our scaling models need to account for this shift. The other thing that could be different in the future is the scale and sophistication of DoS attacks as every new thing connecting to the internet has full communication stacks and decent (or very good) compute/connectivity.

SBA: One area where we expect 5G enabled applications to stress current thinking and drive architectural change is in what I call “Mean-Time-to-Cloud.” There are emerging immersive AR/VR applications, cloud gaming, and concepts around factory floor automation that will require millisecond or less latency between a device that is wirelessly connected at the network edge and the cloud. For the entire 5G infrastructure (RAN+backhaul+WAN) to support that low of a latency requires much more attention to the number of aggregation points, buffer sizes, routing and peering decisions, oversubscription policies, etc. that need to occur between that device and the cloud.We see intelligent service automation, along with lighter-weight IP protocols, and more efficient aggregation techniques as being key changes that are coming to support this. These same comments apply to WiFi 6 by the way. The whole infrastructure, not just the wireless edge has to play a part in delivering the mean-time-to-cloud that the application needs.

JB: Without repeating what was said above, I believe that the last mile carriers will have to adjust their business models as people may shift away from cable/dsl over to 5G hotspots just as people moved away from home phones in lieu of cellphones.  This demand shift will change the ecosystem of how home users function and how carriers build out to support new IoT trends, the ‘more’ connected home, and sustained work from home.

  1. What has been the most challenging part of this pandemic for your company, and how have you handled it?

VGC: coping with operation from home, figuring out how to configure return to work in the absence of an assured vaccine, dealing with revenue variation depending on industries most affected by anti-COVID-19 measures. Dealing with the spread of misinformation, disinformation has also been a challenge in content delivery properties (search, YouTube at Google).

JB: There are two things that we have seen as challenges: 1) Supply Chain – there have been extended lead times from chip makers through switch suppliers which have created a challenge to build out well ahead of our organic growth demand.  Having said that, we are not challenged given the forced downturn of our platform usage and can build out with the supply on-hand until the suppliers catch up; 2) Feet on the Ground – we are having to schedule days in the datacenter vs. daily build/deploy.  This isn’t a blocker, but certainly impedes our ability to move quilty as done pre-COVID19.

  1. Do you think that the Internet is strong and resilient to handle such events in the future? Any changes, improvements you’d recommend we need to make to enhance the resiliency/diversity?

VGC: For the most part, it has held up remarkably well. Need more symmetric access; more capacity; improvement in residential WiFi, routers and firewalls.  Increased redundancy would be helpful. Inter-exchange capacity at ISP peering points would also help.

SBA: Absolutely agree with Vint. It has held up quite well! In addition to the significant (>10,000x) improvement  we’ve been able to drive in fiber capacity we are also seeing how the WebScale operators often think differently about capacity planning. They tend to design their infrastructures to be dynamic and want them to be able to handle peak shopping days or the launch of a new streaming service, or the global downloads of apps or OS updates. This ability to make cloud platforms more programmable, increasing the intelligence of the entire connect, compute, & store infrastructure; we see that as also critical to continuing to increase resiliency.

JB: As mentioned during the webinar, in order for companies and the Internet to expand with redundancy, resiliency, and reliability, we as an industry need to be more open with one another (without compromise to competition) to provide companies with the ability to properly build/scale their networks/interconnectivity with knowledge of capacity, bandwidth, and side of the road accuracy for placement of infrastructure in all cities, states, regions.  This will in turn allow consumers to make the proper design decisions for today and the foreseeable future.

  1. Would one of the panelists or Vint please address the utilization of blockchain technology, specifically how ethereum-based protocols can help with internet security, secure supply chain risk management, etc.

VGC: I am not the right guy to answer as I tend to be quite skeptical of blockchain solutions, especially anonymous ones. I worry about transaction rate processing capacity. I worry about how long you have to store the entire blockchain for reference. It can be helpful to maintain the integrity of digital records and identifying things like date/time of transactions. But you can record false data in a blockchain. It is “just” a digitally-signed, distributed database. Not magic. Nor am I impressed by cyber currencies that have no intrinsic value. Nor am I excited about the instant nature of transfers unless surrounded by other protections like claw-back, etc.