2023-DevDays:NSO in the Brownfield:Orchestrating T-HR Croatia Services

[Nicklas Wagerth]

NSO Customer Success Story and Opening Keynote

Automation Developer Days, Morning Sessions, May 9, 2023 

Abstract

This is a story on how Verso Altima partnered with T-HR Croatia to orchestrate complex VPN service in an expansive MPLS network containing Cisco, Huawei, and Juniper equipment. The session will give you an overview of the entire process, from convincing engineering teams that this is the way to go, to managing expectations, developing models, and reconciling existing configurations, transforming them into simple universal services. We'll show you the true power of orchestration, demonstrated on tricky 'magical' services, such as a single-command transition from VPLS to EVPN. Above all, you will learn how to achieve the transformation to orchestration using reasonable, manageable, and cost-effective steps.

Link to recording

About the Speakers

Goran Košarić, T-HR Croatia

Goran is the IP Transport Group PM leader and an experienced Intellectual Property Specialist with a demonstrated history of +20 Years of working in the telecommunications industry in IP/Transport area. He is skilled in TCP/IP, VoIP, Fixed and Mobile Telecommunications Systems (IP/MPLS, Optical and MW transport, Synchronization, 5G, new Telecom products Development), Broadband 3Play services, Fixed/Mobile Customer Devices. Work experience in people Management, Projects Lead in large cross-functional Teams environments, big Fixed and Mobile Telecommunications infrastructure implementations, new Products and study Case Modelling, Technical and Finance Operational, and Strategy plans development, new HW and SW Platforms Tools introduced into big Operator Networks. Strong legal professional with a Master of Science (M.Sc.EE) focused in Telecommunications from FER Zagreb

Drazen Kedmenec, Verso Altima d.o.o

Drazen is a Senior Technology Advisor with more than 20 years of experience working with service providers, helping them plan, implement, operate, troubleshot, audit, and upgrade complex networks and services. He is especially interested in the end-to-end view of the service provider networks, including work on IPTV, soft switches, turnkey solutions, and IoT. Lately, he’s interested in orchestration and automation, helping service providers move from the traditional CLI approach to incredibly powerful, but challenging end-to-end orchestration. This is a fun way to leverage the accumulated expertise for the benefit of service provider customers.

[Nicklas Wagerth]

Questions and Answers

Q1. What was your most important lesson using NSO? A1. NSO is transformative, on all layers. Getting the teams to start looking at their infrastructure in terms of end-to-end services, not in terms of interfaces, VLANS, addresses, VRFs and such, is very rewarding. Also, NSO is powerful beyond expectations.

Q2. Very interesting presentation! I wonder about the expert/non-expert mode. Do you have any insights on how they were implemented? RBAC? Stacked services? A2. Something close to stacked services, but not quite there. We found out that stacked services are good for some implementations, mostly when everything is done by machines. Since we decided to start with humans, implementing stacked services introduced complexities that can be handled by careful workflow control, which we could not expect from human operators. In other words, we would have to start with a nice web page and a more-or-less complicated workflow manager, and not with NSO CLI. In our models, expert and non-expert modes are implemented within the same service structure (YANG definition), by carefully modeling sensible defaults, simple commands that cover multiple different aspects of service configuration, and finally allowing detailed configuration options if needed. There are also knobs within service models that guide the user towards specific configuration goals (allowing or hiding command tree structures). For example, you can configure L3VPN by simply providing name, device, interface, VLAN, and IP address. The model calculates everything else. Or you can force VRF names, distinguisher, route-targets, descriptions, object names, etc. Of course, everything works pretty much the same regardless of device type.

Q3. Very well presented, thanks! What is the layer you are going to use on the top of NSO? To manage the services or resources Workflows. Thanks A3. We are working on this point intensely, but have not committed to a workflow manager yet (two potential solutions are in testing phase, testing with our service models). The most important point in this regard is that we decided that we will not use NSO for workflow management tasks, although it could be theoretically (and practically) used. Our models are complex internally, but simple to use. We don’t want to complicate things beyond reasonable measure. Also, we want to use a solution that is on-par with the magic of NSO.

Q4. How long time did it take you to develop the first service model? A4. There is a learning curve. Six months from testing the first model to L3VPN in production. After that, one major change in the model structure to make it much more scalable and increase performance a hundredfold. One of the most praiseworthy things about NSO is that it forgives mistakes, even big ones. There are multiple challenges with service modeling. The simplest one is learning YANG and all of the modeling structures available there. For this purpose, NSO NEDs are the best learning tool, covering 95 percent of what you will need when creating your own models. The hardest part is starting to think like a service modeler, not like an operational engineer. You must combine a deep understanding of how things work end-to-end, the implementation details (i.e. commands) and finally how to present the service configuration to human (or machine) operator. Explaining the power of service modeling to your customers is very challenging. “Seeing-is-believing” is the only method that works.