ipSpace.net highlights ML data center design podcast

Ivan Pepelnjak of ipSpace.net flagged a December 3, 2025 podcast episode on the 0x02 Jeffs networking show, where Petr Lapukhov, co-author of RFC 7938 and former Facebook engineer now at NVIDIA, discusses AI/ML data center design. The post highlights Lapukhov's move to NVIDIA and the renewed engineering focus on low-latency interconnects and GPU-to-GPU traffic patterns for AI workloads.

ipSpace.net highlights ML data center design podcast Ivan Pepelnjak of ipSpace.net flagged a podcast episode in a December 3, 2025 post, recommending a conversation in which Petr Lapukhov - co-author of RFC 7938 on BGP routing in large-scale data centers - joins the 0x02 Jeffs networking show to discuss AI/ML data center design. The ipSpace.net post notes that Lapukhov, who was at Facebook for over a decade, has moved to NVIDIA , which Pepelnjak suggests may shape the podcast content. The post is a brief recommendation pointer; the full technical discussion is in the linked video recording. AI/ML workloads have renewed engineering focus on data center fabric design, particularly around low-latency interconnects and GPU-to-GPU traffic patterns that differ from traditional web-tier topologies. What happened ipSpace.net published a short recommendation post on December 3, 2025, flagging a podcast episode featuring Petr Lapukhov as guest on the 0x02 Jeffs networking show to discuss AI/ML data center design, per ipSpace.net. The post credits Lapukhov as co-author of RFC 7938 BGP routing in large-scale data centers and reports he has moved from Facebook - where he spent over a decade - to NVIDIA . Ivan Pepelnjak, the ipSpace.net author, notes this employer shift may influence the podcast content. Technical context Conversations between hyperscale networking veterans and ML infrastructure practitioners typically cover low-latency fabrics, GPU-to-GPU east-west traffic patterns, RDMA-like transports, and switch buffering requirements for distributed training workloads. These considerations differ substantially from traditional web-tier topologies. BGP-based leaf-spine designs and ECMP routing experience from large-scale data center work - the focus of RFC 7938 - are increasingly applied to AI cluster interconnects as GPU pod sizes scale. Industry context Cross-discipline dialogues between routing engineers and ML infrastructure practitioners are valuable for teams designing rack and pod-level GPU clusters. The ipSpace.net post is a two-sentence recommendation pointer, not a technical summary; the full discussion is in the linked podcast recording. Lapukhov's move to NVIDIA, whose networking division Spectrum-X and InfiniBand targets exactly these AI fabric requirements, adds credibility to the conversation. What to watch For the detailed technical content from this episode, watch the full 0x02 Jeffs podcast recording. Practitioners designing AI data center fabrics can also track: - •published slides or transcripts from the 0x02 Jeffs channel - •NVIDIA engineering posts on AI data center fabric and Spectrum-X design - •updates to BGP or RDMA transport standards for ML cluster networking Scoring Rationale A brief two-sentence recommendation pointer from a credible networking blog ipSpace.net by Ivan Pepelnjak, CCIE 1354 to a podcast episode with a notable guest. The topic - AI/ML data center networking with an RFC 7938 co-author now at NVIDIA - is directly relevant to ML infrastructure practitioners, but the primary source provides almost no technical content itself. Score of 3.5 reflects the marginal depth of the post, not the credibility of the underlying subject. Practice with real Streaming & Media data 90 SQL & Python problems · 15 industry datasets Active Users in Target CountriesEasy /problems/sql/active-users-in-target-countries-streaming High-Rated Titles with ReviewsMedium /problems/sql/high-rated-titles-with-reviews User Churn Risk AssessmentHard /problems/sql/user-churn-risk-assessment 250 free problems · No credit card See all Streaming & Media problems /problems/datasets/streaming