Human Infrastructure 371: Follow the Money

Along with guest stars Repatriation, AI Robots, and AutoCon2 (selling out fast).

AuthorAuthor

Ethan Banks & Drew Conry-Murray
September 26, 2024

FOLLOW THE MONEY šŸ¤‘

In this article about whatā€™s really going with return to office, the author points out that RTO isnā€™t about executives needing to control people. Itā€™s not about reclaiming productivity from lazy people who work from home.

Itā€™s about money. Specifically, commercial real estate (CRE). Office towers cost a lot of money to build, and thereā€™s a magnificent mountain of debt outstanding in the CRE market coming due over the next few years. With lots of debt comes lots of pressure on the financial institutions that loaned the money. If they donā€™t get paid back, thereā€™s a real possibility of financial institution collapse with cascading effects on the global economy. So get your butt back into your fabric covered box.

Even though youā€™re the victim, RTO isnā€™t about you. Itā€™s about money. Herein lies a lesson explaining any counterintuitive behavior exhibited by your favorite employer or equipment vendor. Businesses make decisions based on the bottom line, for that is in their nature.

  • Will this proposed acquisition find its way through regulatory approval? Follow the money.

  • Will networking vendors get together and agree on a standardized interface for automation? Follow the money.

  • Will OpenAI remain a non-profit? Follow the money.

Admittedly, this is a cynical way to view the world. Whatā€™s more, it isnā€™t always true. Altruism still exists in the world, as does honesty and integrity. Sometimes people behave in surprising ways. But most of the time, following the money predicts an outcome or explains an otherwise irrational situation. - Ethan

THIS WEEKā€™S MUST-READ BLOGS šŸ¤“

Daniel Dib illustrates from a Cisco perspective how to perform a PAT from a specific source host to a specific destination only. Other traffic from this specific source host should NOT be subject to the PAT. A straightforward configuration, but if youā€™ve ever spent quality time with ip nat { inside | outside } configs, you know they can get hairy. - Ethan

Christian Scholz has not yet slain this dragon. A third fail, a testament to just how challenging this exam is. Should he give it a fourth try? Let him know. - Ethan

Gian Paolo Boarina demonstrates how to programmatically list the temperature of transceivers in a Cisco box running NX-OS. He uses the baked-in Python interpreter to parse CLI output at first. Then he shifts to using NX-OS provided JSON, which intuitively would have been easier to deal with programmatically (because structured data). Nope. The JSON provided by NX-OS was inconsistentā€”sometimes a list, sometimes notā€”making his first approach a bit better in this case. Fascinating. - Ethan

Dmytro Shypovalov presents a primer on Egress Peer Engineering, a way to control how traffic leaves your AS when traditional BGP policies canā€™t get the job done. Dmytro recommends you use Segment Routing over MPLS, although he present alternatives and tradeoffs if SR-MPLS isnā€™t the direction youā€™d like to go. As always, Dmytroā€™s post is clearly illustrated and supported by additional resources if youā€™d like to dig into the topic more deeply. - Ethan

Alkira recently announced an expansion into the Google Cloud Platform (GCP) Marketplace, offering a more efficient acquisition process for GCP customers, enabling them to potentially retire Google committed use dollars while simplifying vendor procurement. GCP customers can deploy Alkira directly within their GCP environment for faster deployment and reduced time-to-value. This streamlined access is particularly crucial as the growing demands of AI workloads are making speed and efficient resource allocation more critical than ever for enterprises leveraging cloud infrastructure. Learn more about the announcement here.

TECH NEWS šŸ“£

This light ā€˜nā€™ fluffy piece on prompt engineering doesnā€™t really argue what the title posits. Instead, the piece is mostly about explaining the fundamentals of how to interact with an LLM to get the sort of result youā€™re looking for. That said, there is something to this article. Prompt engineering is a thing that I believe might be useful for engineers who are incorporating AIOps into their daily work. If youā€™ve got a model trained with IT data you care about, knowing how to ask the questions to get back the answers youā€™re looking for from the model will be valuable.

At the same time, thereā€™s a voice in the back of my head asking, ā€œBut shouldnā€™t the AI grasp what Iā€™m asking for without me having to carefully engineer a prompt?ā€ Because at some pointā€¦isnā€™t engineering a prompt justā€¦coding? AIOps needs to make my life reliably easier, and not become a source of information I have to babysit because I canā€™t trust what itā€™s telling me. - Ethan

This piece reads like an anti-cloud manifesto, taking a steaming dump on all the potential negatives of housing your infrastructure in the public cloud. The piece railed on costs, underutilization, security vulnerabilities, performance & reliability, and vendor lock-in.

Wow. Everyone stand down. If I had to guess, AWS, Azure, and GCP all operate their data centers better than any of could dream of operating our own. They have both the money and the talent to create infrastructure most us simply canā€™t. In addition, if you set aside the cost issues, the problems I see highlighted in the piece are the same problems Iā€™ve seen in any data center Iā€™ve ever worked on.

Repatriating workloads isnā€™t changing the issues of peak load thinking leading to chronic underutilization, endless CVEs, and performance & reliability issues. Only now, those problems are entirely yours to solve. Ohā€¦and your devs still want a cloud-like deployment experience.

Iā€™m not saying repatriating workloads isnā€™t a thing (it is). Iā€™m not saying there arenā€™t significant cost savings to be realized in some situations (there are). Iā€™m saying the hate heading towards public cloud providers lately demonstrates a forgetfulness of the way things used to be when we all ran our own data centers. Everything is a tradeoff. - Ethan

This piece seems to be some noise from AWS to help dispel the notion that they are a monopoly, and fair enough.

If you take this piece together with the piece from Techopedia above, thereā€™s a bit of good news for those of you reading this. If companies are indeed repatriating, then they need infrastructure professionals with very particular skillsets. Networking. Security. Storage. Architecture. Business continuity planning. Private cloud. DevOps. Etc.

Sure, thatā€™s largely what was required for public cloud, right? But now, itā€™s taking the public cloud skillsets and redeploying them to on-premises. More to the point, thereā€™s less being abstracted away (AWS is broken, we dunno) and more deep knowledge required (the K8s cluster is down, weā€™re figuring it out right now, we donā€™t think itā€™s DNS this time, crap it was DNS again) by businesses to keep the infrastructure running.

FOR THE LULZ šŸ¤£

RESEARCH & RESOURCES šŸ“’

This page provides a handy reference for different network automation tools. It breaks out the tooling landscape into open source and enterprise, and arranges tools by categories including network state, telemetry, automation, and others. Itā€™s a useful reference if you ever feel overwhelmed by all your options. - Drew

NIST is requesting comments on the latest draft revisions of its digital identity guidelines. - Drew

Learn IPv6 with this book by Tiziano Tofoni, now in English. - Ethan

Instructor Ed Harmoush strips back cryptography to the very basics to get you started with this daunting knowledge domain that shows up in networking all the time. Edā€™s a great teacher, both personable and clear. - Ethan

3xCCIE, CCDE, DevNet Expert and bootcamp instructor Andreas Baekdahl shares a collection of resources for folks pursuing Cisco DevNet certifications. - Ethan

From the README.md. ā€œThis is an open list of web crawlers associated with AI companies and the training of LLMs to block. We encourage you to contribute to and implement this list on your own site. See information about the listed crawlers and the FAQ.ā€ - Ethan

AutoCon2 is coming up fast on November 18-22 in Denver, Colorado. The AC2 team has been sharing the reg numbers with us, and this one is going to sell out. Soon. If you want to be there, register ASAP. Full stop.

Workshop Registration (classes before the conference) opened August 8

  • Thereā€™s a a great slate of workshop options covering a range of topics in network automation and orchestration

Not sure if you want to go? The full AC2 conference agenda has been published to help you make up your mind.

AutoCon is THE forum for Network Automation. We at Packet Pushers will be there covering the event, and hope to meet many of you in person. See you in Denver!

INDUSTRY BLOGS & VENDOR ANNOUNCEMENTS šŸ’¬ 

The Selector AI product for network operations continues to iterate. In this announcement, Selector highlights the following.

  • A ā€œNetwork Language Modelā€ (NLM), which they claim is an industry first. The details in the release are scanty, but what I think is happening here is that Selector has trained a model on networking telemetry. Now, you can ask that model natural language questions to get information about whatā€™s going on in the network.

  • Improved digital twin capabilities that should enable additional interesting ā€œwhat-ifā€ scenarios. Modeling like this is valuable assuming you can think of the what-if questions to ask. Sometimes, what-if scenarios exist because someone on the team had A Bad Experienceā„¢, and they want to avoid ever having that experience again.

  • Programmable synthetic sensors. Sounds like application performance monitoring on the surface of it. Iā€™ll assume thereā€™s some AI baked in there to help correlate network telemetry with application events. Therefore, if the synthetic test came back with a poor number for an app, Selector would be able to tell you that the poor performance was due to a network problem.

Packet Pushers will be talking to Selector.AI in coming weeks to fill in the blanks. Stay tuned. - Ethan

If youā€™d like to influence the future direction of IETF standards related to network management protocols, youā€™ve got an opportunity. Act soon, though.

ā€œA workshop organized by the Internet Architecture Board (IAB) aims to chart a path for the development of future network management protocols and techniques. The Next Era of Network Management Operations (NEMOPS) workshop will begin by assessing the impacts of the previous IAB workshop on both network operations and protocol development.ā€

ā€œWorkshop participation will be by invitation based on ā€˜position paperā€™ or ā€˜expression of interestā€™ submissions, though even those not planning to participate are invited to submit position papers on workshop topics. Submissions are due by 16 October 2024, with full details available on the NEMOPS workshop webpage.

The workshop will be held online, 3-5 December 2024, and anyone can sign up for the [email protected] mailing list to keep track of further announcements and workshop-related discussions.ā€ - Ethan

Allegroā€™s multimeters arenā€™t the things youā€™re thinking of if you grew up tinkering with electronics like I did. These are network analysis devices that would be part of your network analysis and troubleshooting toolkit.

In this release, Allegro announced their new 1010 and 3010 multimeters. Both models are ā€œequipped with 2 x SFP28 ports with hardware-based timestamps for 1 / 10 / 25 G links and 3 x 10 / 100 / 1000Base-T ports. The number of ports can be increased by 4 via adding a network expansion card, with expansion capabilities up to 400 Gbit/s.ā€

I think ā€œhardware-based timestampsā€ is a key feature here. You probably know if youā€™re working in an environment where youā€™d benefit from that level of precision. More info about these devices on the data sheet. - Ethan

Commvault buys Clumio. This gives Commvault reach into AWS S3 data protection. Iā€™ll be virtually attending & covering Commvault Shift in October 2024, where I expect there will be more noise about this acquisition. - Ethan

The components required to get us to 1.6Tb Ethernet are coming. Case in point, Broadcom has announced its Sian2 DSP PHY, which provides 200G lanes of electrical and optical interfaces. From the press release: ā€œSian2 DSPs enable pluggable modules with 200G/lane interfaces that are foundational to connect next generation AI clusters.ā€ Broadcom anticipates full production of the Sian2 by the end of 2024, with Ethernet switches using the PHYs to roll out in mid-2025. - Drew 

SonicWall has released a Threat Brief that focuses specifically on attacks against the healthcare sector. The report finds that ransomware was involved in 91% of attacks. More than half of the vulnerabilities exploited by attackers were used against Microsoft Exchange. You can read the full report here, but be warned; it might make you ill. - Drew

TOO MANY LINKS WOULD NEVER BE ENOUGH šŸ³

  1. Why Tech Employment is Not What It Used to Be - Crepuscular Circus

  2. Itā€™s Okay to Abandon Things - NetNinja

  3. Things Iā€™ve learned serving on the board of the Python Software Foundation - Simon Willisonā€™s Weblog

  4. A terrible way to jump into colocating your own stuff - Rachel By The Bay

  5. http:, ftp:, and ... dict:? - Terence Edenā€™s Blog

  6. How to Show Which Ansible Config Changes Are Being Applied - Packet Coders

  7. SR Linux Containers Run on Apple Silicon - ipSpace

  8. The LinkedIn AI saga shows us the need for EU-like privacy regulations - TechRadarPro

  9. Explaining Soperator, Nebiusā€™ open-source Kubernetes operator for Slurm - Nebius Blog

  10. Portnox Announces Support for Microsoft External Authentication Methods (EAM) - Portnox Press Releases

  11. Comcast is Harnessing Leading-Edge Cloud and AI Tech To Transform the Way Its Network Delivers Next-Generation Internet Experiences - Comcast Press Releases

  12. Why Comcast Uses DriveNets Network Cloud - DriveNets Blog

LAST LAUGH šŸ˜†