Human Infrastructure 370: Leading a Team, AI Uncertainties, RTO Mandates

AuthorAuthor

Ethan Banks & Drew Conry-Murray
September 19, 2024

LINKEDIN USES YOUR DATA FOR AI TRAINING - OPT OUT

In the Packet Pushers community Slack (everyone welcome), Frank Seesink (thank you!) shared a warning that LinkedIn has opted all of us in to use our data to train their AI. If you’d rather LinkedIn didn’t use your data to train their AI, you should opt out. In the browser, that setting change is easy enough.

Settings > Data Privacy > How LinkedIn uses your data > Data for Generative AI Improvement > Use my data for training content creation AI models > OFF.

I am not a fan of huge companies using my data to train their AIs. I should have a unique right to my intellectual property, similar to the protections I get from US copyright law. I believe that, eventually, privacy and other ethical abuses related to training AIs will be outlawed. For now, technology has mostly outpaced the ability of governmental bodies to act. It’s up to us to limit how our data is used as best as we can. - Ethan

TAKE OUR SHORT & HILARIOUS AUDIENCE SURVEY

Okay, our audience survey is not that hilarious. It is a little funny, as we ask some questions designed to make you smile, laugh, chortle, guffaw, snicker, teehee, lol, smirk, or…uh…I’m out of synonyms. Taking the survey won’t take much of your time, and it helps us out as we set direction for our independent little enterprise. Over 6,000 of you sub to this newsletter. If all of you filled out the survey, that would be amazing. That is, don’t leave it to someone else to fill it out for you. There’s only one you, and your thoughts are unique—the stats we gather are nice, but we value your comments the most. THANK YOU. - Ethan

THIS WEEK’S MUST-READ BLOGS 🤓

Andrej Stender goes deep on Linux nftables, which you can use for packet filtering. Many of us have used iptables or nftables for simple packet filtering. But what happens when your packet filtering needs become complex? What happens when you’re doing lots of networking functions on your Linux box? Andrej has the answers. He dug through the docs, read sundry guides, experimented on his own, and created this guide so you don’t have to. - Ethan

Péter addresses being an engineering manager when your company is on a corporate wartime footing. I found myself conflicted reading through his recommendations. Péter’s advice centers primarily around the welfare of the company. Survival. Making sure that the company makes it through the tough times by rapidly shipping the products that will keep the lights on. I get that. I’ve been that guy, leaning as hard as I dared into my team to deliver on the projects and billables driving cash flow. But I think there’s a more critical aspect to corporate wartime than the company itself Péter somewhat minimizes—people.

Companies come and companies go. More of them will go than survive. Sacrificing yourself to a company so that it survives is likely to burn you to the ground, whether the company ultimately makes it or not. Is that worth it? Maybe, but in my experience, almost certainly not. Even in today’s climate of tech people being laid off (our best to those of you cut in the latest Cisco attrition), your mental health, ability to be present for loved ones, and spiritual well-being all matter. If your company is on a war-time footing and they’ve placed you on the front lines, it could be that another job is better for you.

I know, I know. If you can find another job. It’s tough out there right now. - Ethan

Brad Casemore articulates the mental exercise many of us have been going through regarding generative AI. Considering the massive computational cost, is there a business ROI? If there’s no ROI, what’s the point of GenAI to decision makers considering the bottom line? To the first question, Brad’s not sure. He’s leaving the ROI calculations up to the investors with the assumption that if the ROI doesn’t show up, the investments will start drying up.

As to the second question, that’s harder to answer. It could be that GenAI is indeed a step on the path to the transformative technology of the future—artificial general intelligence. That’s useful if so, but to me, GenAI and AGI are fundamentally different technologies, even if they feel similar. It’s not clear to me how GenAI techniques would be used by a future AGI. The future of GenAI? Murky, I think. - Ethan

Rasika Nayanajith reviews a Bluetooth serial adapter from Intertooth. He likes it. No more looking for drivers for a USB-to-serial. The Intertooth devices has some other slick features, too, such as writing the device config automatically when you type “exit”. - Ethan

Author Toni Pasanen has published this book introduction as well as Chapter 1. Math ahead…you’ve been warned. - Ethan

NetOrca

With NetOrca you can allow your internal customers to consume all the great automation infrastructure you have created and manage it throughout its service lifecycle. See our demo with Ethan here: https://youtu.be/fPBF9Dsoloc?si=YvvRzeLFbyEPWy3X

TECH NEWS 📣

Are return-to-office mandates effective? The deck, or subhead of this article from the Sloan Management School, says everything you need to know: “Your organization’s highest-performing employees want executives to focus on outcomes and accountability, not office badge swipes.” Instead, executives at major tech companies (most recently Amazon) want to go back to what this article calls “management-by-monitoring.” The article also says that’s the least-effective management method. Companies that claim to be driven by innovation and data sure do seem eager to move backwards towards a work policy based on face time rather than any meaningful productivity metrics. To my mind this movement is driven by a need to exert control more than anything else. - Drew

The Wi-Fi HaLow standard is 802.11ah, and it’s aimed at low power and long distance applications for IoT sensor networks. HaLow operates in frequencies below 1Ghz, which helps with the distance requirement. In general, the lower the frequency, the further you can go with it. The higher the frequency, the more subject to attenuation the signal is.

The Wi-Fi Alliance says HaLow is good for about 1 kilometer (see graphic above). Morse Micro has pushed beyond that spec right to the theoretical limit in this test in Joshua Tree National Park that followed up another long distance test on San Francisco beach.

“The throughput of HaLow in this new testing topped out at 2 Megabits per second at 9.9 miles — likely a best-case scenario considering how little interference there is in a rural area like Joshua Tree National Park compared to a place as densely-populated as a San Francisco beach.”

Sure, 2Mbps doesn’t sound like much, but there are many use cases that fit comfortably into that much available throughput. - Ethan

Intel’s Foundry company will be making AI custom chips for AWS, a multi-billion dollar deal. That’s a big win, and highlights the emphasis Pat Gelsinger is placing on Intel Foundry in Intel’s turnaround plan. Check out the Intel Newsroom release later in our newsletter for more info on that. - Ethan

FOR THE LULZ 🤣

RESEARCH & RESOURCES 📒

The NERD Project - The Network Engineer Resource for Development - Tony Bourke via GitHub
https://github.com/tonybourke/Project-NERD

Instructor Tony Bourke (frequent podcast guest) released this resource to help with network automation labbing. He says, “What is project NERD? It's a collection of tutorials and accompanying artifacts (files, playbooks, scripts, etc.) to build certain types of VMs and containers that aid in network automation, specifically labbing of network automation. The instructions are geared towards home labs, but they can be used as the basis for production builds. These toolkits should provide an environment to explore network automation with containerlab, VS Code (as a web app), and Ansible (and/or Python).” - Ethan

INDUSTRY BLOGS & VENDOR ANNOUNCEMENTS 💬 

Itential has released version 1.1 of their free torero tool, which automatically creates execution environments for scripts and playbooks. The big announcement here is about clustering. Why cluster your torero environment? Wyatt Sullivan explains…

“Where I would cluster is when I have any environment where script run times are important, there’s more than one person running them, and we need to get consistent responses. Environments such as:

>_ CI/CD Pipelines: multiple builds happen at any given time.
>_ Support Ops: uptime is key.
>_ Monitoring tooling.
>_ Deployment tooling.
>_ CIO is watching.

Any of these situations will have important markers that need the network scripting to consistently work in a known, reliable amount of time.” - Ethan

Nokia has launched data center automation software that aims to compete with the likes of Juniper Apstra and Cisco ACI. At present, this automation software only works with Nokia’s SR-Linux network OS, but the company says it will add support for Cisco, Arista, and SONiC soon.  I wrote about the announcement here if you want more details. - Drew

Netris makes software (a controller and agents) that lets you operate on-prem network gear as if it were public cloud infrastructure. It ties in with DevOps tools such as Terraform and Kubernetes to turn network capabilities into services or resources that other applications can consume. Netris recently released a new version of their software that supports Nvida’s Spectrum-X Ethernet switches and Nvidia Bluefield NICs. From the press release: “Netris switch-fabric management functionality for NVIDIA Spectrum switches is designed to automate the day-0, day-1, and day-2 phases of switch fabric operations.” - Drew

The IT and OT realms seem to be getting closer. For folks unfamiliar with industrial and building control networks, this post provides some useful background on BBMDs, or BACnet Broadcast Management Devices. BACnet uses broadcast messages to communicate among devices, which can cause problems on IP networks. BBMDs are used for “translating BACnet packets into a format that won’t get blocked by an IP router, and back into a format that BACnet devices need to communicate.” There’s more in this post if you’re interested. - Drew

Intel CEO Pat sent a note to the troops letting them know the state of the company. As most Intel news in the last year or more has ranged from negative to “meh”, some reassurance was likely in order. Pat highlighted 3 major initiatives.

  1. “We must build on our momentum in Foundry as we near the launch of Intel 18A and drive greater capital efficiency across this part of our business.

  2. We must continue acting with urgency to create a more competitive cost structure and deliver the $10B in savings target we announced last month.

  3. We must refocus on our strong x86 franchise as we drive our AI strategy while streamlining our product portfolio in service to Intel customers and partners.”

Pat’s note goes on to explain how they’ll execute on these goals. Interesting reading if you’ve followed Intel for any length of time. They’ve got an uphill battle, although I wouldn’t classify it as sisyphean. - Ethan

Source of truth and a great deal more, the OpsMill team has released Infrahub version 0.16. Highlights include…

  • “Resource pool assignment for numbers (integers such as ASNs) and IP addresses/prefixes.

  • Improved Git repository state visibility, logging and notifications.

  • Increased speed and visibility of Proposed Change diffs between branches”

Infrahub is currently an open source offering in open beta. Also…otters. - Ethan

  1. Mastering IP Routing: The Vital Force Behind Modern Connectivity (YouTube) - Wendell Odom’s Network Upskill

  2. My New SR-MPLS Course, and Krzysztof’s SRv6 Blogs - Network Fun Times

  3. Looking for 240/4 Addresses - Potaroo

  4. Code Completion in VS Code with Ollama - Josh-V

  5. Windows DHCP Server on AWS – Creating Static IPv6 Addresses in AWS - This Bridge Is The Root

  6. CloudZero’s Cloud Efficiency Rate Metric - Ethan Banks via LinkedIn

  7. CIQ Launches 'Ledger Pro,' an All-in-One Solution that Provides Deep Visibility Into IT Infrastructure, Automations - Cision PRWeb

  8. Global telecom leaders join forces to redefine the industry with network APIs - PR Newswire

  9. Qumulo Delivers World’s First Cloud Native Unstructured Data System - Qumulo Press Releases

  10. Arelion Launches New Route from Monterrey to Querétaro, Expanding Network in Mexico - PR Newswire

  11. Keysight Introduces Advanced 14-bit Precision Oscilloscope for General Applications - Businesswire

LAST LAUGH 😆