distributed-systems.
15 writings found
Latest Archives
How Cloudflare Reclaimed 100TB of RAM with Better Hashing
Deep dive into consistent hashing optimization: how understanding the math behind algorithm design led to massive memory savings at scale.
How Cloudflare Reclaimed 100TB of RAM Through Better Hashing
Cloudflare optimized consistent hashing in Pingora, reducing memory footprint by 25% through struct alignment fixes and mathematical analysis of hash distribution.
ZGateway: The Proxy That Controls Uncontrollable Scale
How Meta's ZGateway proxy layer solved the connection mesh problem at billion-operation scale by moving complexity from clients to infrastructure.
Why Proxies Win at Scale: ZGateway's Lessons for Platform Teams
How Meta's ZGateway proxy tier solved the connection mesh problem at scale, and what this teaches us about infrastructure design choices.
How Meta's ZGateway Proxy Solves the Million-Client Problem
Understanding how a shared proxy tier manages connection meshes, enables cross-client batching, and scales reliability at hyperscale infrastructure.
The Proxy Layer Pattern: Where Infrastructure Gets Smart
How Meta's ZGateway shows why interposing a managed tier between clients and backends solves problems no individual client library can.
Why Proxies Win at Scale: Learning from Meta's ZGateway
How interposing a stateless proxy tier between millions of clients and a shared backend solves reliability and efficiency problems that client libraries cannot.
MetaRoCE: Rethinking Network Transport for AI at Scale
Meta's new MetaRoCE protocol redesigns Ethernet transport for million-GPU clusters, prioritizing edge intelligence over fabric control. What this means for infrastructure engineers.
Aurora DSQL Goes Global: What Multi-Region Consistency Means for You
Aurora DSQL expands to 16 regions with multi-region clusters. I break down why active-active consistency matters for distributed systems and what it changes for developers.
etcd v3.7.0 Released: RangeStream, v2 Store Removal and More
etcd v3.7.0 ships RangeStream RPCs, drops the legacy v2 store bootstrap, overhauls protobuf deps, and brings real CPU savings for Kubernetes clusters.
Cloudflare Meerkat: Building Consensus Without a Leader
Cloudflare's new Meerkat service uses QuePaxa to eliminate leader-dependent consensus bottlenecks across 330+ global data centers. Here's why it matters.
Meta's Adaptive Ranking Model: The Real Cost of Serving Trillion-Parameter Ads
Meta scaled ads recommendations to LLM complexity while keeping latency under a second. Here's why their inference trilemma solution matters beyond advertising.
Meta's RCCLX: Why AMD's GPU Communication Stack Just Got Interesting
Meta open-sources RCCLX with Direct Data Access and FP8 collectives for AMD GPUs. A deep look at what this means for multi-GPU AI workloads.
Meta's RCCLX: Why AMD GPU Communication Just Got Interesting
Meta open-sources RCCLX with Direct Data Access and low-precision collectives, potentially reshaping distributed AI workloads on AMD hardware.
Meta Open Sources RCCLX: AMD Gets Serious Performance Boosts for AI Workloads
Meta's RCCLX brings Direct Data Access and low-precision collectives to AMD GPUs, delivering 10-50% speedups for LLM inference on MI300X hardware.