Andrew Zuo·Sep 9iPhone Duo? Please. The Mac Studio Is The Biggest Apple September-ish AnnouncementI don’t understand folding phones. I’ve never understood folding phones. You buy a phone that is twice as expensive as a normal flagship…A response icon6A response icon6
Goutham Kalla·Sep 3Ever Wonder Why RDMA Doesn’t Make Your Database Faster?You have a distributed transaction that takes 60 milliseconds to commit. You profile it. Most of the time is “network.” So you do the…
Shivank Chaudhary·Aug 26From 16 GB/s to 189 GB/s: Debugging a 400G RoCE Fabric for NCCLA systematic walk through everything that can silently throttle GPU collective performance — MTU enums, NUMA topology, PCIe ACS, Kubernetes…
Goutham Kalla·Aug 14Your Transaction Is Waiting on the Speed of LightWhy cross-datacenter commits are slow, why the usual fixes don’t help, and what happens when you stop waiting for consensus before you…
Shriom Tripathi·Aug 8Two-Sided RDMA Strikes Back: Why Disaggregated Memory Databases Are Rethinking the One-Sided…The Core Insight: The one-sided versus two-sided RDMA argument was never really about bypassing the remote CPU. It was about key…
Gautam Vasudeva·Aug 7RDMA — The Secret Network Behind AI Supercomputers [Article 4]RDMA From Scratch — Speaking the Language of the NIC
Gautam Vasudeva·Jul 30RDMA — The Secret Network Behind AI Supercomputers [Article 2]How a NIC Can Talk to Memory
Abhishek Mishra·Jul 27How a Packet Travels from a NIC into GPU Memory: GPUDirect RDMA ExplainedStory so far, kind of :)
