NANOG 85 Agenda

Full Abstract

The NANOG 85 Hackathon will focus on the theme of Projects that Make a Difference.
Learn more at https://www.nanog.org/events/nanog-85-hackathon/

On Wednesday, May 25th, we will hold the Hackathon welcome, introduction, infrastructure tutorial, idea-pitching, and team-forming session over Zoom; this session will be recorded. Saturday, June 4th, will be the start of the hackathon; this day will be all virtual regardless whether or not you are at the conference venue. Sunday, June 5th, will be a true hybrid day with people continuing to work virtually as well as dedicated facilities (workspace, wifi, etc) for those at the conference venue.

The Hackathon starts with a brief welcome and introduction, tutorial, and team formation on Wednesday, May 25, at 1:00pm Eastern. Hacking begins virtually at 1:00pm Eastern, Saturday, June 4. The hacking ends at 5:00pm Eastern, Sunday, June 5, when the team presentations will begin.The Hackathon will conclude around 6:00pm Eastern Sunday, June 5. We have dedicated Support/Help Hours on Saturday from 1:00pm - 4:00pm Eastern virtually via Zoom and again on Sunday from 12:00pm - 5:00pm Eastern in a hybrid format.

Full Abstract

Sylvie will speak to us about developing the Internet during a pandemic and experiences building a submarine cable during a pandemic (all without travels...)

Full Abstract

In 2016, the conclusion of a study conducted in 22,000 publicly traded company in 91 countries by The Peterson Institute for International Economics, revealed that the ones with women in leadership roles had reported 6% higher profit. While this could be attributed to several factors, one cannot underestimate the unique values those determined and hard-working leaders brought to their industries. Fast-forward to 2022 as the world prepares to embrace more relaxed COVID-related restrictions, opportunities such as working remotely are further expected to pave the way for a higher percentage of women to enter the job market.

This panel will discuss the facets of leadership, challenges, successes, and opportunities.

Full Abstract

Join us for the Women in Technology Lunch and hear a profound case one of the biggest personal challenge that is common to Women. A 10 minute "real talk" (fear of being vulnerability defied) will explore a real life example of its amplified impact on a woman in the tech industry. This will challenge you to reflect, maybe it resonates with your experience (woman, man or other gender identity); or contemplate your positions of strength to help others along.

It is in these moments of vulnerability that we get to pull each other in, pull each other up, break biases (self-impose or systemic) and empower every woman in Tech.

Full Abstract

Operating a hyperscale network at scale reliably - abstract/simulate/emulate/validate:
1. Abstraction through compartmentalization and self-contained routing building blocks
2. Simulation vs emulation - when and why
3. Validation through assertion - continues validation and comparison of intended and operational states.

Full Abstract

Metro Ethernet networks have evolved from pure layer 2 architectures based on commodity switching to IP/MPLS architectures that are more scalable and reliable, but also more costly and complex. We discuss an emerging architectural approach that builds on low cost, commodity, disaggregated switching to create a scalable, resilient IP underlay and a simplified, automated service delivery overlay based on VXLAN and SDN control. We also describe how this architecture has been deployed in a US-based broadband Internet Service Provider (ISP) network.

Full Abstract

The ‘edge’ is an overloaded word in the industry, and can mean several different things to different stakeholders, depending on the industry and use case. The main purpose of the edge is to bring compute and storage closer to where is needed, providing lower latency and distributed processing. Important benefits for applications like Autonomous Vehicles, Gaming and Healthcare. Switches not only need to connect the servers within the edge, but also need to connect the edge to the central Datacenters, potentially those two functions collapse in the same switch .
Kubernetes is the preferable cloud management solution at the Edge and can also serve as a small and nimble network management system: By design Kubernetes can be used to manage the network: A key characteristic of Kubernetes is that its true declarative API can be extended with ‘custom resource’ that can be used to encode domain knowledge for specific applications. The Kubernetes declarative API can then be used to configure the network and enforce policies.
On this session we will bring an introduction how could you take advantage of GNMI/YANG standards, to manage network devices, and Kubernetes CRDs, and automate the network requirements in the underlay network. We'll bring some examples of use cases that can show the great potential of those technologies combined.

Full Abstract

There are a limited number of IPv4 addresses in existence—only 4.3 billion—and IPv4 exhaustion has become a major issue in the industry as the world is slow to adopt IPv6. For a new buyer or seller entering the market, it can become overwhelming to understand the basic process of transferring IPv4 space, protecting themselves from fraud, and ensuring their businesses receive resources on the right delivery time table. During this panel we’ll be hearing from a couple of experienced participants in the IPv4 market that can expand on mitigating transaction risks for buyers and sellers, evaluating pricing in an opaque market, using an IP address sale as an opportunity to upgrade your network, and navigating ARIN's transfer process.

Full Abstract

In October 2019, I presented “Some boring network engineering interview questions and how to replace them with smart ones” at NANOG 77 in Austin, TX (https://www.youtube.com/watch?v=g35UumfZ-H4&t=941s). An attempt to challenge the status quo around the quality of interview processes that was a new and different topic at the time. The recording of the session on YouTube has had one of the highest number of views and the authors received numerous questions and positive feedback. This time for NANOG 85, we are submitting a follow-up topic "The Anatomy of the Most Challenging Network Engineering Interview Question".

Full Abstract

Efficient analysis and collection of deeply inspected, high throughput network traffic is hard… especially as the trend towards globally distributed deployments continues.

pktvisor (https://pktvisor.dev) is a free and open source analytics agent designed to address this challenge. It can analyze flow, packet capture, dnstap and other inputs, combining deep traffic analysis with data sketch algorithms to efficiently extract counts, top-k heavy hitters, set cardinality, quantiles and other key information from data streams directly on the edge. The result is lightweight time series metrics that plug into modern observability stacks.

Orb (https://getorb.io) is a related open source project that acts as a control tower for a distributed fleet of agents, providing fleet and configuration management along with data collection and sinking functionality, accessible via web UI and REST API.

Together their goal is to deliver immediately actionable insights local to the traffic source and simultaneously collected and integrated into global result sets.

This talk will introduce the origin of the tools, discuss the goals and status of the projects, and look to the future as they extend beyond traffic analysis and into general network debugging and analytics embedded at the edge.

Full Abstract

Don’t miss our Community Meeting for an opportunity to hear about what is happening with NANOG and the Program Committee.

Full Abstract

The Domain Name System (DNS) has evolved to include new, longer top-level domain (TLD) names, as well as TLDs and email addresses in various languages and scripts. The Universal Acceptance (UA) of these Internet identifiers means that they work seamlessly with all Internet-enabled applications, devices, and systems, some of which include email tools, webservers, content management systems, and more. Achieving UA-readiness is necessary for providing broader Internet access to global end users and driving digital inclusivity. Join the session to learn about the challenges and solutions for making UA a reality, including the gaps that need to be addressed by the technical community.

Full Abstract

Everyone is doing RPKI nowadays, but ... how does one troubleshoot issues within RPKI itself?

Getting RPKI Route Origin Validation going is straightforward: install a validator and point the EBGP routers to the RTR service endpoints. But what kind of files is the validator really pulling in? How? How to decode the binary formatted files into something human readable? What kind of references do the files contain towards each other? What the heck even is a "Signed Object"?

In this presentation I'll share an overview of the data structures contained inside ROAs, GBRs, and BGPsec Router Keys, and explain how to use BSD/Linux utilities to inspect this type of data.

Full Abstract

NANOG 85 Hackathon Recap

Full Abstract

Following up on interest in learning how other NOGs work, we're bringing together five NOGs in a panel to give a quick overview of their Modus Operandi and general philosophy.

Starting with a brief intro from NANOG, the format is short ~5 minute NOG presentations followed by a panel discussion around the benefits of the different models especially in the age of Covid. We would like to get audience interaction driving the Q&A section.

The NOGs represneted on the panel are:
NLNOG, Netherlands
SiNOG, Slovenia
SAFNOG, Southern Africa
UKNOF, UK
DENOG, Germany
ENOG, Eurasian

Full Abstract

ECMP is a key in today’s IP network. It provides both efficiency and resilience. However, the detecting and troubleshooting of these ECMP paths remains a big challenge that faces many operators. FIB corruption or FIB misconfiguration at any node is one among many issues that can happen and may lead packets to follow a non ECMP Path.
Path Tracing is a new solution that reveals all the unknows about ECMP. It allows operator to detect the currently available ECMP paths. They can also detect the ECMP Paths that were available at a given time in the past. Operators will be able also to know whether packets are following an ECMP Path or not. In addition, they can detect where packets are being drifted from the ECMP Path. Moreover, Path Tracing provides operators accurate end-to-end delay, per-hop delay, and the load of each interface that forwards the packet along the path.
The Path Tracing dataplane is optimized for HW linerate implementation in the base HW forwarding pipeline. It requires minimum packet header parsing and processing. As a result, it can be implemented using a minimum number of micro code instructions.
Path Tracing has also a rich ecosystem that includes several implementations in merchant silicon (Broadcom, Cisco, Marvell, others) and open source (Linux, VPP, P4, others).
The solution is currently being standardized in the SPRING WG at IETF (https://datatracker.ietf.org/doc/draft-filsfils-spring-path-tracing/).
In this presentation, we will review the path tracing solution and its ecosystem

Full Abstract

Optical fiber carries over 95% of terrestrial internet and private network traffic, and over 99% of international traffic via undersea cables. It is a miraculous medium, but it has fundamental limits in terms of the combination of capacity and reach. With almost 5 billion kilometres of optical fiber already installed, there is a strong incentive to extract as much capacity from existing fiber as possible.
In this tutorial we explain what those fundamental limits are, and what we can do to get as close as we can to them based on a hierarchy of lowest impact to highest impact on the network:
- The lowest impact implies using both the existing optical fibers in the ground today, and the existing line systems that are installed on those fibers.
- Involving more investment would be to use existing fibers, but to enhance and upgrade the line systems.
- And the highest investment option would be to deploy new fiber types.

We will explain each of these options, including giving an explanation of emerging fiber technologies. We will also attempt to show the potential increase in capacity from each of these options.

Full Abstract

Like many tasks, draining traffic off core routing device should be a simple task, but as you dive deeper into it (depending on the architecture) can become a complicated task with many moving parts to track and tasks to complete. This talk is going to walk through the process I went through to document, understand, automate, and improve the task of draining traffic off a routing device to be able to perform maintenances more often with low impact to customers. This talk is is both about the tool which was created but also about the process, and the iterative nature to get the tool to the state it's in now.

Full Abstract

Power costs involve both consumption and provisioning. Traditional approaches to provisioning in large systems (e.g., Cisco ASR 9000, NCS 5500, Juniper MX/PTX, etc) tend to overestimate power requirements and push operators toward populating more power supplies than they really need. This is exacerbated by the wide range of optics power in the 400G generation (up to 800W / RU), which break traditional static approaches to power budgeting. Improving provisioning can save money (fewer PSUs and power feeds), free up facility resources (e.g., consume less cooling budget), and still prevent brownouts. It can also improve the efficiency of PSUs (they are most efficient at 30-100% load so too many PSUs can waste power).

My experience is in leading enhancements to address this for Cisco's 8000 series, but the concepts apply to many platforms and operators will benefit if they are applied more broadly.

In this talk, I'd like to encourage operators to revisit this area of sw/facility interaction as a way to explore efficiency improvements.