Introduction and Welcome

• Keith Mitchell (UKNOF)

Room: Charter 1, 2 & 3

Dropping in 80Gbits (sort of) of Firewalls with OpenBSD (Main Session)

• G Llewellyn (Myself)

Room: Charter 1, 2 & 3 Managing a network that did stateless filtering at the edge leaving stateful, granular firewalling to the hosts _(100% Linux)_ was all well and good till the business went for SOC2 compliance... Firewalls had to be dropped onto the edge where there was ~60Gbits of transit capability but it had to be done without any disruption to the various IP streams coming into, out of and between the data centers. Being a "startup" this was a unique opportunity to leverage OpenBSD _(pf, OpenOSPFd and OpenBGPd)_ to move fast but _not_ break things _(and save a tonne of money at the same time!)_

5G - the next era of mobile (Main Session)

• Neil McRae (BT)

Room: Charter 1, 2 & 3 This will discuss 5G, and the what, the why, and the when!

Sponsor Presentation: Building for Ultrafast (Sponsor-led Content)

• Andy Furnell (Zen Internet)

Room: Charter 1, 2 & 3 We think the future is ultrafast. Since 1995, Zen Internet has been at the forefront of the UK Internet’s evolutions. And right now, we see no slowdown in increasingly complex, data hungry applications that rely on fast robust connectivity. Whether this is businesses migrating to cloud computing and hosted voice or consumers’ depending on connectivity to underpin smart homes and a fundamentally connected life. The last year has seen 40% growth in peak average internet traffic. The networks upon which our lives and businesses increasingly rely have never been so important. Performance issues or drops in service – can take an axe to a company’s profits or ruin an evening of catch-up TV! That’s why we think ultrafast, ultra-reliable networks are the future, and it’s why we built one – in just six months. Andy will detail why we decided that a major network upgrade was necessary for both Zen’s customers and partners, and how we built a next-generation network – from scratch – in ultrafast time. He’ll talk about Zen’s simultaneous on-net rollout, and how an ambitious programme of Local Loop Unbundling has given us the fifth largest network in the UK. He’ll put it all together to explain why networks need to be fit for a data hungry future.

None of us knew what we were doing, we made it up as we went along - Part 2 (Main Session)

• Paul Thornton (PRT Systems Ltd)

Room: Charter 1, 2 & 3 Some more amusing anecdotes, historical artifacts, and possibly interesting technical tales of woe from the UK Internet of the late 1990s.

IPv6 networking on Arduino (Lightning Talks)

• Nicholas Humfrey (BBC)

Room: Charter 1, 2 & 3 IPv6 is starting to become more commonly supplied by major ISPs in the UK and has been well supported for a long time by mobile phones and desktop operating systems. The percentage of data being transported over IPv6 (compared to IPv4) is expected to go up enormously in 2017 within the UK. However many embedded, niche and hobby devices are a long way from supporting IPv6. Because IPv6 addresses are 4x longer than IPv4 addresses, it is a common assumption that it also uses a lot more resources as well. I wanted to challenge that assumption and see if it was possible to implement IPv6 on a very constrained device, such as Arduino. My implementation has support for SLAAC, DNS Client, a basic HTTP Server and UDP Client and Server. Example programmes fit within 2kB of RAM and 16KB of Programme Memory. While the implementation would not pass ipv6ready certification, it is at least as functional as a similar libraries for IPv4. Writing an IPv6 library for Arduino was is a personal challenge that allowed me to explore IPv6 in a lot of depth and understand the limits of the Arduino platform better.

PGP Key Signing

• Harry Reeder

Rise of the IoT Attack Vector

• Sean Newman (Sponsor)

Room: Charter 1, 2 & 3 There’s nothing new about Botnets, but one built from IoT devices is – and, they are taking the scale of the DDoS threat to a level not previously anticipated. Understand why this new vector is on the rise and the detail behind the types of attacks it is delivering.

FD.io: building bespoke software data plane network functions. (Main Session)

• Maciek Konstantynowicz (Cisco)

Room: Charter 1, 2 & 3 FD.io is a Project at Linux Foundation focusing on a modularised and extensible packet processing software framework for building bespoke network data plane applications. It is a collection of projects centred around VPP - Vector Packet Processing - a high-throughput, low-latency and resource-efficient data plane services platform. This talk introduces FD.io, covers the VPP architecture, current main integration efforts and key performance benchmarks. It then explores applicability to network deployments, use cases and shows a path to one terabit software router (1TFR).

SDN, CMDB, NMS… CRM?! How we're putting the customer at the centre of our network. (Main Session)

• Marek Isalski (Faelix Limited)

Room: Charter 1, 2 & 3 Following on from some of the ideas presented in Brian Nisbet's talk "Your 60 seconds starts now", my talk would be to explain how we put our Customer Relationship Management system right in the middle of our network infrastructure. The primary audience of the presentation is smaller ISPs/NSPs/etc, who might not yet have the ITIL-style business processes in place, or are just thinking about Configuration Management Databases; but it might give some ideas to some larger providers as well. Estimate: ~20-30 minutes. Format: some talking around bullet points, some "lolcat GIFs" for levity, but also some screenshots of system in use, and examples how we're scripting and using APIs to build our SDN/CRM/CMDB/NMS alphabetty-spaghetti-monster. Intro and Motivation: Faelix is a small "boutique" provider, small number of staff, emphasis for our customers is quick access to "technical support with clue". That kind of support is expensive to the business (no first-line helpdesk). And like anyone else in the room, we have to respond quickly: to potential customers, to billing enquiries, to service problems, etc. We're doing many of the usual network automations already, taking inspiration from e.g. Facebook's talk at UKNOF32 (of course, we're not that scale!). How can we make the less-technical parts of our business processes more automated so we scale better, more like a "tech startup", i.e. so we don't need 10x staff to get 10x customers? Execution: We've put CRM in the middle of everything. Some slides explaining how e.g. * customer VM doesn't get spun up unless a customer exists in the CRM, the VM exists, and it has availability/capacity/etc monitoring profiles applied * the customer/VM/etc record configures the network (routers, switches, VMs, etc) * which means the customer is being billed (hurray!) * and we have a record of what they're being billed for (don't accidentally switch off live things) * and their payments are coming in (cashflow is king) * reducing accidental (or on purpose) freebies (anecdotal problem for small hosting providers; and means we have quantitative figures for the "hosting donations" we make) * we know what $customer_vm is connected to, interacts with... i.e. everything to rebuild their service (CMDB!) * makes maintenance notifications easy, because we can grab all customers on a particular device and email newsletter them; and potentially see they've read it * and makes marketing new services to customers easy too (as long as CRM does good email newsletters) * and start to ask customers for feedback for our own improvement Further Ideas: We're starting (and probably by UKNOF37 will have finished) getting our Network Monitoring System interfaced with our CRM so that monitoring events in Icinga2 are reflected on customer records in CRM. This will mean we could automatically apply service level credits to customers' accounts (assuming we have such an outage!), but also means we have one place to look for any customer-affecting issues: their record in our CRM has everything. i.e. full customer life-cycle within the CRM. Conclusions: Some examples/anecdotes about how this is helping us, basically "all the info we need for customer service management, visible in one place". Link back to beginning: if something is taking longer than Brian's 60 seconds, and the customer calls us up, we don't have to go scrabbling around for information in different silos of systems (and we don't have to keep multiple systems up to date with the same data).

Large BGP Communities and Shutdown Communication. (Main Session)

• David Freedman (Claranet)

Room: Charter 1, 2 & 3 Both "Large BGP communities" (*RFC8092*) and "BGP Shutdown Communication" (*draft-ietf-idr-shutdown*) are new standards designed to improve the lives of operators of modern ISP networks. In this talk we explore these and discuss adoption.

BGP Session Culling (Lightning Talks)

• Will Hargrave (UKNOF/LONAP)

Room: Charter 1, 2 & 3 With this BCP Internet-Draft we hope to draw some attention to good practises which can be applied by IP networks or IXPs to mitigate negative impact caused by maintenance operations on lower layer networks. The idea is to promote the concept of breaking the control-plane in a controlled fashion, before actually breaking the data-plane. This is a follow-on from my lightning talk at RIPE67. Since then the idea gained popularity and is applied at more and more IXPs. The video + pdf are available here: https://ripe67.ripe.net/archives/video/116/

OARC's Software Tools Suite for DNS Data Capture, Analysis and Sharing (Main Session)

• Keith Mitchell (UKNOF)

Room: Charter 1, 2 & 3 The Domain Name System Operations Analysis and Research Center (DNS-OARC) is a non-profit, membership, organization that seeks to improve the security, stability, and understanding of the Internet's DNS infrastructure. This presentation will include a brief introduction to DNS-OARC, and an overview of OARC's extensive suite of software tools for DNS data gathering, sharing and analysis, including recent efforts to refresh, consolidate, update and extend this toolset to modern development platforms and practices. Recent and upcoming OARC Internet-wide data capturing activities will also be discussed.

Continuous Delivery for Network Engineers (Main Session)

• Simon Gunton (Mr)

Room: Charter 1, 2 & 3 Simon has been at AutoTrader for over three years and has witnessed a business change from traditional waterfall to fully agile. I will giveyou a whistle stop tour of what continuous delivery is and, how some of those methods that sysadmins have been using for while can now start to also be used in a networks role.

BIRD route-server configuration: click, done! (Main Session)

• Diego Neto (NL-ix)

Room: Charter 1, 2 & 3 One of the missions of an internet exchange is to promote and facilitate peering between organizations: this operation is generally performed via route-servers. Unfortunately, at the moment, this approach comes with a few severe limitations that are not encouraging organizations to join route-servers: there is no possibility to easily decide who to peer or who to peer not within the connected members, plus the management of the route servers is usually something obscure, not standardized and still performed in a really manual way, that means inefficient, error prone and time consuming. On top of it, in case of a peering network distributed within different locations, also latency becomes a key factor in the selection of your peers and the filter mechanism based on community strings, once again, turns out to be definitely limited. NL-ix decided to find a solution to the above-mentioned situation and implemented a solution on top of the open-source route server BIRD. What came out is a complete automated system that gives to the members full and fine-grained control on their peering sessions via a friendly web interface or - due to his modular design – via API/WebHooks at each level of the automation chain. The system, using exclusively open-source technologies (Python, PostgresSQL, Django, brocade StackStorm , Flask), has been realized in collaboration with Brocade and allows to go over all the traditional limitations, allowing filtering and prepending connected members based on latency, datacenter location and specific IPs; once your configuration is done it gets converted in a BIRD configuration file, pushed and safely deployed on the route-server; everything gets logged and is trackable plus the extra possibility to easily integrate the automation platform with instant messaging software to keep the operation department updated about the outcome of the process.

NetMcr: Networking for Networkers in Manchester (Lightning Talks)

• Tom Hill (Bytemark Hosting)

Room: Charter 1, 2 & 3 - Basic introduction to NetMcr & format - Why we started doing it - What's happened to date - How to get involved - Please spread the word - Thanks