Introduction and Welcome

• Keith Mitchell (UKNOF)

Room: Charter 1, 2 & 3

The Single Source of Truth for Automatic Network Configuration (Main Session)

• Andy Davidson (Asteroid)

Room: Charter 1, 2 & 3 A presentation about how a single source of truth, expressed in an elegant data model, is used to operate an Internet business' process and network automation. Many automation presentations to date have considered programming techniques/skills/languages a network engineer embarking upon an automation project shall need. Or, concentrated on a vendor's automation features, so that the audience can see the Arista or the Juniper integration options. Little has been produced to date which explains how an engineer will integrate software relevant business processes or product design. If an IXP (but equally an ISP, a hosting company, etc.) concentrates only on the automation platform facing their network infrastructure, whilst the instruction set to manage the network is automated, without integration into the company's products or customer's requirements, can the company really be said to be automated? When Asteroid embarked upon a platform that could build and operate fully autonomous peering platforms, it became clear that the automation systems that we build must have a deep integration with the network switches, and the servers that will support the platform, but also the business processes that would be used to create and operate exchanges/port services. When a company extends the scope of the automation project into the product set, sales process, monitoring there are a number of efficiencies realized: - Freedom to provide services by nontechnical teams - The speed of deployment of customer services (reduce time to bill!) - The accuracy of monitoring systems - More customer self-service options - Rich API that customers can deploy into their own software - SLA and outage validation **Presentation to cover:** A technical presentation that explains key concepts/("lessons learned") to networking companies (ISPs, IXPs, content companies) looking to embark upon an automation project. Concentrating specifically on: - Why and how to build a data model that can describe your customers, products, and network, teams - What normalization is, and why/how to use it - Why and how to abstract different layers of technical systems to allow vendor changes/flexibility - How and why to use the data model to build systems configurations and monitoring templates - How and why to abstract between technical elements (like "ports") and all matters relating to the service on those technical elements - How and why to expose parts of it to customers to provide an extra layer of transparency and benefit to your end users - How to integrate with data which is in third-party databases - The mistakes I made and had to refactor out after launch

Event-driven network automation and orchestration (Main Session)

• Tom Strickx (Cloudflare)
• Mircea Ulinic (Cloudflare)

Room: Charter 1, 2 & 3 One of the major challenges in networking is the diversity, in terms of data representation, which is often vendor-specific. Vendors APIs are inconsistent and incomplete, some mainstreams platforms are closed, and custom software is not allowed on your device. By combining Salt proxy minions, with third party libraries such as NAPALM, which presents the data in a vendor-agnostic shape, we are able to leverage the DevOps methodologies in networking. NAPALM support is now integrated in the official Salt releases, beginning with Carbon and improved in Nitrogen. Beyond cross-vendor configuration management, reaction to internal and external network events becomes easy and there are no orchestration boundaries.

Measuring uptime at IXPs and NIS Directive (Main Session)

• Robert Lister (LONAP Ltd)

Room: Charter 1, 2 & 3 Governments all over the EU are looking at how to implement "The Directive on security of network and information systems (NIS Directive)" and IXPs are wondering how this will impact them. A key part of the NIS Directive is reporting outages and availability. Given the IXPs limited perspective and point of view in the Peering landscape are such availability figures useful? Should IXPs adopt a standardised approach? This presentation explores various approaches to monitoring uptime and availability at IXPs. Can IXPs ever have a meaningful "99.999% uptime" figure? Do we keep it simple, or does combining multiple factors in the uptime calculation make the metric over complicated? Are there any standard approaches IXPs could consider?

IXPDB Update (Lightning Talks)

• Andy Davidson (LONAP / 2Connect)

Room: Charter 1, 2 & 3 IXPs are considered core Internet infrastructure and carry a growing percentage of the world's Internet traffic. While there are a number of disparate websites and resources, which contain various data on IXPs, none are globally authoritative or fully comprehensive. This makes it difficult for network operators and other interested parties to make informed decisions related to the global Internet ecosystem. The Internet eXchange Federation (IX-F), aims to develop a system that will fill this gap. Acquiring data directly from IXPs through automation, it will aggregate and promote useful third-party data sources, provide valuable insight, reporting, data export and visualisation functionality. The aim is to work closely with the Internet technical community throughout the project and host workshops that help to promote use of the system and encourage the application of best common practices (BCPs). The presentation will provide an overview of the database and our future plans.

Network Configuration Automation at LINX (Sponsor-led Content)

• Riccardo Verzeni (LINX)

Room: Charter 1, 2 & 3

PGP Key Signing

• Harry Reeder (N/A)

Room: Foyer

No, Bob, "The Cloud" is not the answer (Main Session)

• Chris Malton (Individual)

Room: Charter 1, 2 & 3 * What is "The Cloud" - At least in the context of this presentation. * What "The Cloud" is good for * Why it doesn't work for small to medium businesses. * What was this client running? * Why was it expensive? * How I planned and delivered the move.... * .... and made it better than Amazon * .... and made it cost less than any "Cloud" offering * How open source software made high-availability easy

GDPR and the Internet: Evolution not revolution (Main Session)

• Jonathan Langley (ICO)

Room: Charter 1, 2 & 3 This talk is intended to give an overview of the requirements of the GDPR as it applies to the internet, and will look to explain that the GDPR is an evolution of existing data protection legislation. It will cover important definitions that are particularly relevant to internet companies, and will also cover our soon to be updated cloud computing guidance, specifically covering the new requirements on data processors.

How We Found a Firewall Vendor Bug Using Teleport as a Bastion Jump Host (Main Session)

• David Farrar (Exa Networks)
• Marek Isalski (Faelix Limited)

Room: Charter 1, 2 & 3 First part of the talk: what is Teleport, how do you use it, what might the use-cases be in a NOC environment. Second part of the talk: how Exa Networks used Teleport and a bunch of Raspberry Pis to find a serious performance problem in the default config of one particular vendor's firewalls.

Memcached - DDoS Moves into the Multi-Terabit Era (Sponsor-led Content)

• Sean Newman (Sponsor)

Room: Charter 1, 2 & 3 Although cybercriminals continue to use IoT botnet-powered DDoS to great effect, recent months have seen the rise of a completely different and, to date, much more deadly DDoS attack mechanism. This latest innovation from the cybercriminal community sees the commonplace memcached general-purpose caching system being used to deliver DDoS attacks in the multi-terabit per second range. Understand why this new vector requires an always-on approach to DDoS mitigation and what hosting providers can do to avoid being part of the problem

IPv6 in an office (Main Session)

• Tim Chown (Jisc)
• Tim Bray (ProVu Communications Ltd)

Room: Charter 1, 2 & 3 This talk explores the challenges and problems when using IPv6 in a small business. Based on 10 years of my own experience. We will explore some of the failover options available for IPv6 users. How IPv4 nat makes changing ISPs easy. And how some of the ideas coming out of the ITEF are ok in theory, but probably not that great in practice.

Scaling for Ultrafast, G.FAST, FTTP, 5G and the Cloud. (Main Session)

• Neil McRae (BT)

Room: Charter 1, 2 & 3 Presentation will describe BT's network development over the past year to scale up to Ultrafast speeds in the core network, deployment of BT's Network Cloud with MPLS o/UDP, SDN and deployment of the latest high capacity routing platform enabling 400G interfaces and SR, Together with NFVI and 5G core development Driving the latest in telemetric network management and closed loop automation. We will share some interesting traffic and utilisation stats and trends.

Kea DHCP - a modern DHCP Server (Main Session)

• Tomek Mrugalski (Internet Systems Consortium)

Room: Charter 1, 2 & 3 Kea DHCP is a modern Open Source standards-compliant DHCPv4 & DHCPv6 server that runs on Linux, BSD, and MacOS (just like ISC DHCP). Kea is production-stable and under active development. Hear what functionality is available now, what's coming soon, and what direction the roadmap is likely to be taking (feedback welcome).

Notworking - Untangling the mess (Main Session)

• Chris Malton (Individual)

Room: Charter 1, 2 & 3 This short presentation covers several incidents I ended up managing while in my role at Vostron in Southampton. It covers: * The ticket which was resolved with an internal recommendation to hire a cat. * The curious incident of the missing backup phone line. * When "OK" means silently corrupting data. * How a "Road Traffic Accident" leads to 12 weeks of downtime. 4 short "war-stories" type tales that network & system admins across the country can probably relate to in one way or another.