Networking Field Day and Nokia's EDA
TL;DR
诺基亚 's new DC network automation product, Event Driven Automation (EDA), is an impressive new product, especially for a v1. It uses Kubernetes under the hood but you don't need to understand Kubernetes to use the system. Plans for multi-vendor support and brownfield deployments will broaden the environments where it can be used. In my view, EDA is a strong entry in the DC network automation field, and if you operate DC networks, EDA is worth your time to investigate.
The Deets
On 24 Sep 2024, I had the privilege of being part of a group of delegates with Tech Field Day that spent the day with Nokia for a deep dive on their new DC automation product, EDA.
The Broader Network Automation Perspective
In the big picture, I continue to be excited and optimistic about what's happening in network automation overall - people are hungry to learn and do, and there is a growing set of tools from open source, vendors, and interesting integrations among and between them. This has been emphasized in great detail over the last 18-24 months for me via my perspective in the Network Automation Forum . Nokia is adding a solid contribution to this growing automation ecosystem with EDA. EDA is continued validation that better network automation matters and is desired, in the DC and beyond.
To k8s or not to k8s?
During the event, there was initial emphasis on use of Kubernetes under the hood of EDA, but that users don't need to understand Kubernetes to use EDA.
TBH, this was really off-putting for me at first; coming from a largely "trad netops" perspective, I was overwhelmed with the thought of having to spin up on yet another set of tools from the software world just to do what we've done for decades without another cool new technology: just operate the network. And while it was made clear that you didn't need to be k8s-literate to use it (and reinforced by demos and lab exercises), there was muscle memory throughout the day about the advantages of k8s for this environment and tooling.
To be sure, there are audiences that will be attracted to EDA’s use of k8s constructs - ones that are very cloud- and software-centric. Other audiences (smaller DCs and networks, less software expertise, more "trad netops" shops) won't fully appreciate those internals and may also be turned off a bit.
By way of personal confession, I've been avoiding going beyond a surface understanding of k8s for awhile now, and as I hear the advice I've given others about the disruptive nature of tech to our jobs (and yes I hear you 迈克尔 ), it's probably time for me to take my own medicine and learn more about k8s and its applicability to network automation and its broader uses. I'm not saying I'm ready to go all in on k8s, but I know it's time for me to personally go deeper to make better assessments. This might be true for you too.
There is a great opportunity here for Nokia and other entities in the industry to educate netops personnel on k8s concepts for networking, starting with very simple concepts first; I think (hope?) that this will be part of what Nokia is planning in one of the workshops at the upcoming #AutoCon2 conference.
The Demos
The EDA demos were very effective at helping us understand how the tool is intended to be used. The first thing that struck me was the emphasis on managing DC operations with a great UI and options for providing fabirc details: most options had forms to fill in architecture info, along with making the yaml available if you want to edit the text yourself.?
There are many options to sort through, but exploring those options during the labs wasn't hard at all. Some guidance and more time spent on workflow concepts would be helpful in future versions. The volume of automations to create device configs was impressive - it simply just worked and eliminated lots of keyboard activity.
The AI-powered natural language interface was also very impressive; you can use it to ask questions about operational stats, device and system state, calculate summarized statistics, and much more. Using the natural language interface also generated the query language derived from the questions asked and instructions given, providing both a "tutor mode" to help you see and learn how the queries are constructed, while also allowing you to grab the query language statements and fine tune those queries if you can't get your natural language query just right.
What did you Intend?
In both declaring info for a fabric definition and asking questions via the natural language interface, I noticed some practical issues around what it means to declare intent; the details matter and there is no single standard for this (and I'm good with that), so I think we'll continue to see different approaches to declaring intent across products coming out in the next few years. Being clear about intent will require art and science, and will perhaps become netops' version of Google hacking and prompt engineering; this will be interesting to watch.
The Labs
The labs reinforced the key demo concepts, starting with very explicit instructions that guided us along the way. If you want to investigate EDA, I highly recommend going through these lab exercises.
And while our time going through the labs wasn’t streamed or recorded (I mean, don't you just love watching other people do labs?), they are one of the best ways to see what EDA does, especially those of us who learn best by doing. Adding labs that focus on common real troubleshooting scenarios will also be helpful for people evaluating the product, and will be a must-have for netops teams.
Near Term Wants
The automation ecosystem keeps getting stronger, with infrastructure vendors (including Nokia) making their networking devices more automate-able, using tools from both open source and from vendors that have a high degree of interaction and cooperation.?
EDA supports two network operating systems today, Nokia’s SRlinux and SROS, with plans to add other NOS's in the future. Adding more NOS support is a good thing.
The ability to manage existing Python scripts and use them in workflows will also be helpful for brownfield deployments. Demos and labs showing this will also be helpful to others. In my view, we don’t spend enough time in netops designing workflows, and we don’t think about how automation and other tools can help us design improved workflows. This is something I want to spend more time on; look for future posts and TNOps podcasts on this.
The demos also touched on the extensibility of EDA thru modularity and replaceability of functions - if there are pieces you don't like, there are multiple ways to insert your preferred modules. This is another area that would benefit from more detail on how complex this is to do, and labs that show what's involved would be a great addition for future EDA investigators.
All of these items will help make EDA stronger and more broadly applicable, giving users another strong option for DC network automation and operations - and more choices with good tools is a good thing.
Remaining Questions
These are some of the questions that netops teams will be asking about EDA and competing products:
Of course, cost needs to be included here. And for a tool that is so heavily focused on operational savings, OPEX needs to be the focus of this analysis. This analysis is something we’re generally not comfortable with as engineers and in the industry, but it does need to be a part of an honest evaluation of EDA and similar products.
Wisdom in Many Advisors
This piece is just one view of what we saw on EDA. I highly recommend that you take a look at what the other delegates saw:
And please review the session recordings so you hear what was said for yourself. They are well worth it - you can find them here:
Ultimately, if you think this may be useful in your DC environment, you should definitely contact Nokia for a closer look.
Last but not least, thank you to all the people at Nokia who made this event happen:
And many others I'm sure I missed?
It was well worth the time to hear from you all and learn about EDA.
To Learn More
We'd love your input on going deeper on topics like EDA and DC network automation on the TNOps podcast on Packet Pushers - please check us out here:
Thank you for your thoughtful comments Scott Robohn, both the encouraging words and the advice. You have a unique view of the market, so all you said helps us immensely in delivering our ambitious plans for automation using Nokia EDA. Admittedly, at #NFDxNokia, we spent a lot of time walking through the theory, the investigation of the landscape, before coming up with EDA. That’s what happens when you break the mold and start a revolution :) But we aimed for ease of use to get to the correct outcome: - use of low-code/no-code - minimizing clicks to get what you want done - LLM for quick answers, and teaching moments - no-compromise access to the internals, but careful consideration that most NetOps folks need to get things done quickly I know from your article that you got all those points so I again want to thank you for patiently spending a day with us at Nokia. Ooh and maybe we jogged you into cracking that K8s book open!
Product Line Manager
5 个月At first, I didn't recognize your voice Scott Robohn thanks for the write-up! I will have a lot of fun creating those labs for the public consumption this December. The perspective you shared will help me shape the course appropriately