• Human Infrastructure
  • Posts
  • Human Infrastructure 363: AI Is A Turnoff, A Free Python Course, High Tech Moral Bankruptcy, And More

Human Infrastructure 363: AI Is A Turnoff, A Free Python Course, High Tech Moral Bankruptcy, And More

AuthorAuthor

Ethan Banks & Drew Conry-Murray
August 01, 2024

THIS WEEK’S MUST-READ BLOGS 🤓

IPv6 Buzz podcast host Ed Horley reports from Network Field Day 35 about presenter Hedgehog. What’s Hedgehog created? Well, to hear their presentation, you’d believe the solution to be an automated Ethernet fabric on whitebox switches for AI workloads. Networking for AI is a hot story right now, and Hedgehog is certainly offering that. But Ed sees a bigger picture.

“I think the compelling story is the fact that you can build an Ethernet fabric easily with the same workflow and concepts as what you are used to with a public cloud service like AWS or Azure for networking. This is important for organizations that are trying to streamline their AI workload processes and want to leverage the cost saving of running AI workloads on fixes capital investments versus recurring costs in public cloud. It still leverages VXLAN and EVPN, but it is abstracted away from the operator. The reality is, this is where network automation is going. A controller that builds out Ethernet fabric solutions in a standard way where you don't need to touch or maintain the underlay and an API for those that want to consume the fabric can use to set it up they way they need it with a simple abstraction. Let me repeat that - this is where network automation is going.”

I think Ed’s right, and as he also points out, Hedgehog is not the only entrant in this space. - Ethan

Day Two DevOps podcast host Ned Bellavance offers a guide to migrating from Terraform to OpenTofu. He starts out with a quick recap of why OpenTofu came to be, then jumps into how you’d migrate to OpenTofu if you were so inclined. The initial migration pattern Ned shares is “cautious and methodical” with lots of thoughtful sanity checks along the way—appropriate for production. But he also shares a YOLO migration approach suitable for environments you don’t care so much about.

But should you migrate to OpenTofu? There is no right answer to this one, but Ned concludes with several considerations to help you make the decision. - Ethan

Forrest Brazeal notes that AWS is, somewhat haphazardly, killing off non-core AWS services. That is, products bolted on to the core infrastructure offerings and aimed mostly at developers. Does this mean AWS is the next Google, killing off, for whatever their insidious reasons, things many of us are using? Forrest doesn’t think so.

“I haven’t heard too many complaints from actual users of the services AWS is turning off, because in most cases, there just aren’t any users. Nobody has chosen CloudSearch on purpose in years; it’s long since been replaced by OpenSearch. These services have been in maintenance-only mode for some time.”

Sounds like the list of ridiculous services AWS cert candidates might be expected to name is becoming mercifully shorter. - Ethan

Security researcher Tal Be’ery has a long and highly detailed post about the inner workings of AWS session tokens, which he claims to have reverse-engineered to see hwo they work. Be’ery writes “AWS STS Session Tokens play a vital role in AWS’s security model by enabling the use of temporary, limited-privilege credentials for accessing AWS resources, thereby enhancing AWS’ overall security and access control.”

So why try to peer inside via reverse engineering? Be’ery says AWS doesn’t share many details about its authentication and authorization protocols. He writes “...we believe that it is utterly important for defenders and builders to understand their environment’s Authentication and Authorization protocols and the resulting credentials structure, we need to reverse engineer what the missing parts from AWS official documentation.” - Drew

Ivan Pepelnjak builds this post in response to an online question posted by Daniel Dib (I love these kind of back-and-forth engagements, BTW). Ivan boils down the question to this: “What can you use to implement layer-3-only VPNs?” He then provides very clear background basics, some advice for what he would do, and then a direct answer to the question. (He also throws some elbows at vendor marketing for muddying, rather than clarifying, the issue). - Drew

<RANT> For decades, Silicon Valley leaders have fed us a lot of happy-talk bullshit that their guiding principles are innovation, the betterment of our lives, and not being evil. All of it has been cover for their actual objectives: to build monopoly power as fast and hard as possible; to enrich themselves regardless of any harm to what they consider externalities (users’ mental health, privacy, the environment, truth, etc.); and to block or hamstring any regulations that would hamper the first two goals.

These objectives aren’t exclusive to Silicon Valley. They drive most large corporate sectors (e.g. oil and gas, pharma, finance, food conglomerates). But Silicon Valley has been able to repeatedly charm, trick, or confuse the media and government institutions into believing that tech titans have the public’s best interests at heart. They don’t. Let us act accordingly. </RANT> - Drew

The (US)NUA is a national community of network engineers who are passionate about networking technologies.

Network engineering is a rapidly evolving field with a plethora of new technologies introduced all the time. You may be wondering, how does one keep up with all of this new knowledge? The answer is the (US)NUA!

The goal of the US Networking User Association and its Affiliated Groups is to get network engineers together regularly - throughout the country - to discuss the latest and most relevant topics in the network world, in a vendor agnostic and sales-free environment. 

Our Affiliated Groups hold regular formal get togethers and occasional informal meet-ups around the country. There are also tools for keeping in touch like Slack, Twitter & LinkedIn, offering multiple ways for you to network with our growing base of professional engineers anytime the mood might strike you!

TECH NEWS 📣

Steven J. Vaughan-Nichols reports that the biggest new feature of OpenTofu 1.8 is the ability to use variables and refer to local modules when in 1.7 this was not possible. There’s also a number of other enhancements making OpenTofu a viable Terraform alternative, although the legal drama with HashiCorp continues. - Ethan

Some marketing professors have published a study of consumers that finds that when products are described as using AI, they are less popular than identical products that don’t include AI features in the description. More specifically, the researchers say AI-infused products “decreases purchase intention” and that would-be purchasers have a lower level of “emotional trust” in products that include AI in their descriptions. This research is focused on consumer goods, but I wouldn’t be surprised if you could find a similar effect for IT products and enterprise buyers. The article linked above is a good summary of the research. The paper’s abstract is available here. - Drew

Here’s updated information on a problem we shared an issue or two ago. Supposedly, microcode was going to fix the CPU problem. But alas, it’s not that simple. Intel has stated the following.

“Intel Core 13th and 14th Generation desktop processors with 65W or higher base power – including K/KF/KS and 65W non-K variants – could be affected by the elevated voltages issue. However, this does not mean that all processors listed are (or will be) impacted by the elevated voltages issue.

Intel is confident that the microcode patch will be an effective preventative solution for processors already in service, though validation continues to ensure that scenarios of instability reported to Intel regarding its Core 13th/14th Gen desktop processors are addressed.”

Note the word “preventative” in regards to the microcode patch. If the patch wasn’t deployed before the CPU was damaged, your best best is to RMA the unit. - Ethan

Intel stock is down. They aren’t winning in the AI/GPU market. The anticipated Gelsinger turnaround hasn’t happened yet. The Raptor Lake CPU thing. Next? Reportedly, cutting heads which continues a theme begun in October 2022. More tech industry attrition. Sorry to hear it for all of you affected. - Ethan

The headline is most of the news here. French authorities seem to have a vague idea of who might have done it, but have made no arrests or even firm accusations. Several French providers were impacted, including SFR, Bouygues and Free. Orange was not affected.

Did anyone see a BGP disturbance that maps to this event? The article reports the cuts to have happened around 2:15am CEST on 29-July-2024. - Ethan

FOR THE LULZ 🤣

Shared on Bluesky by @trinastechnobabble.bsky.social

RESEARCH & RESOURCES 📒

Blip is for measuring latency and packet loss. Who cares? You do, as more bandwidth only solves some throughput challenges. Excessive or jittery latency as well as packet loss affects user experience directly. All the bandwidth in the world won’t fix packets never showing up on the other end, or showing up at long or unpredictable intervals.

See Blip in action in your browser here. - Ethan

The description reads, “A guide into the world of Kubernetes Services and Network APIs. In this video, we explore the fundamental concepts of Kubernetes networking, including how pods communicate within a cluster and how external traffic finds its way to the right place. We'll take a closer look at Services as a way to expose your applications, defining ClusterIP, NodePort, LoadBalancer, Ingress, and Gateway API.” About 30 minutes long. Enjoy! - Ethan

Kirk Beyers has launched a free course for network engineers (and anyone else) who wants to learn Python. It’s a ten-week course with lessons sent via email. Did I mention it’s free? - Drew

INDUSTRY BLOGS & VENDOR ANNOUNCEMENTS 💬 

While EC2 Hibernation is not new to the AWS cloud, hibernation of EC2 instances running on AWS Graviton processors is. Hibernation can reduce your cloud bill and save you some time. You don’t have to rebuild the EC2 instance from scratch, plus “While an instance is hibernated, customers pay only for the storage of EBS volumes [where the EC2 instance persists to], including the saved content from the instance memory. There are no charges for instance usage or data transfer during hibernation.” - Ethan

Microsoft has released a technical overview of the outage triggered by a CrowdStrike update that took down millions of Windows devices. The blog provides sample output from a crash dump to dig into the cause of the outage from the Windows perspective. The post also shares “how customers and security vendors can better leverage the integrated security capabilities of Windows for increased security and reliability.” - Drew 

TOO MANY LINKS WOULD NEVER BE ENOUGH 🐳

  1. Is Computing a Discipline in Crisis? - CACM

  2. Surveilling the Masses with Wi-Fi-Based Positioning Systems (16 page academic PDF by Eric Rye & Dave Levin) - arxiv

  3. Access Plex behind CGNAT - Varun Chopra

  4. The internet is in decline – it needs rewilding - The Guardian

  5. Essays on programming I think about a lot - Ben Kuhn

  6. Engineering for Slow Internet. How to minimize user frustration in Antarctica. - brr.fyi

  7. Application Scenarios for the Quantum Internet (RFC 9583) - IETF

  8. How Websites Know You're Lying About Your User-Agent (2021) - Christopher Tarry

  9. How Python Asyncio Works: Recreating it from Scratch - Jacob Padilla

  10. Is regulated BGP security coming? - APNIC

  11. Reporting on Border Gateway Protocol Risk Mitigation Progress; Secure Internet Routing - US Federal Register on behalf of the Federal Communications Commission

LAST LAUGH 😆

How Machine Learning Works

Shared by Eddie with Drew via a Twitter DM