Hardware Podcasts

Systems Software in the Large

Oxide and Friends · 25 Sept 2025 · 1h 44m · Episode 28

What this episode covers

Dave Pacheco, leading Oxide Computer Company's Update project, explains the architectural and organizational challenges of building a full-system, self-service update mechanism for rack-scale computers. He details the use of the 'plan-execute' pattern, 'rendezvous tables', and Dropshot versioning to manage complex state transitions and ensure system reliability during updates. The discussion highlights strategies to combat 'organizational procrastination' and maintain focus on long-term goals through daily water cooler meetings, demo-driven planning, and rigorous scope management.

Most relevant to

large enterprises

Topics

Mentioned

Companies
Oxide Computer Company
Products
Update, Reconfigurator, Nexus
Tools
Dropshot, DTrace, Wireshark
Also
Cockroach

Related episodes from other shows

Other shows on the same ground, closest first.

  • Embedded.fm

    34: Really Big Cabbage

    Elecia White discusses firmware update mechanisms, explaining the risks of bricking devices and the trade-offs between different bootloader architectures. She details strategies for constrained systems, including using external RAM, ping-pong buffers, and security measures like encryption and crypto chips. The conversation covers the 'goat, cabbage, and wolf' puzzle as an analogy for managing update sequences safely. Key technical decisions involve choosing between simple serial updates, complex networked OTA, and secure key management for consumable-based systems.

    Embedded firmwareSecurity & privacy
    7 chapters

  • Coredump Sessions · 57m

    #002: The secrets to building secure & scalable OTA infrastructure with Nick Sinas

    Nick Sinas, CTO at Oval, and Chris Coleman, CTO at Memfault, discuss the critical importance of Over-the-Air (OTA) updates for connected devices, emphasizing that OTA must be designed into the hardware architecture from day one. They detail the technical requirements for secure, scalable OTA systems, including bootloader design, memory management, and cloud-side cohort rollouts to mitigate fleet-wide risks. The conversation highlights real-world failures, such as breaking future update capabilities due to stack overflow fixes, and stresses the need for rigorous testing, security best practices like signing keys, and fallback mechanisms like dual-bank architectures.

    Embedded firmwareSecurity & privacy
    10 chapters

  • Embedded.fm · 1h 12m

    390: Irresponsible At the Time

    Tyler Hoffman, co-founder of Memfault, discusses the challenges of managing firmware updates, device diagnostics, and remote monitoring for millions of IoT devices. He draws on his experience at Pebble and Fitbit to explain why existing cloud platforms are often unsuitable for resource-constrained embedded systems, highlighting the need for secure boot, staged rollouts, and efficient heartbeat reporting. Hoffman details how Memfault’s platform helps engineers detect crashes, track battery life trends, and prioritize bugs by aggregating data from large device fleets. The conversation emphasizes the importance of defensive programming and building robust OTA systems to prevent bricking devices at scale.

    Embedded firmwareConnected products & IoT
    8 chapters

  • Over the Air · 34m

    Revolutionizing IoT Device Management: Updating the World with Peridio

    Bill Brock, COO, and Justin Schneck, CPO, of Peridio, discuss the challenges of managing IoT device fleets and their platform's approach to device-centric CI/CD. They explain why web development abstractions fail in embedded systems, highlighting the critical need for atomic updates to prevent 'bricked' devices from power failures or poor connectivity. The conversation covers the shift from simple sensors to complex edge AI, requiring collaborative workflows that bridge embedded engineers, data scientists, and web developers. They argue that current cloud providers offer scale but lack the necessary collaboration tools and safety nets for physical device distribution.

    Connected products & IoTEmbedded firmware
    7 chapters

  • The Pragmatic Embedded Podcast · 50m

    Agile at Growing Hardware Startups with Matt Cole

    Matt Cole, Director of Embedded Systems at Rise Robotics and formerly of Super Pedestrian, explains how he applied agile principles to embedded software development at a growing hardware startup. He details the challenges of integrating firmware updates for e-scooters, managing parallel hardware and software teams, and navigating supply chain crises that forced reactive engineering. Cole emphasizes the importance of treating the engineering team as a customer to improve tooling and process efficiency while maintaining safety and quality.

    Embedded firmwareScaling production
    7 chapters

The summary, chapters and topics are generated from the episode’s audio by this directory, not written by Oxide and Friends. The episode itself belongs to its publisher.