Loading…
Topic
812 posts about Performance. Every summary links to the original.
See the idea
Bare metal gives engineers direct control over GPU, network, and virtualization layers, but it shifts elasticity and lifecycle work onto the team. This trade-off map shows when that bargain pays off.
Bun packs a runtime, package manager, bundler, and test runner into one binary. Here is how those four parts fit together and where the seams still show.
A five-step how-to for deciding when an AMD Instinct GPU should stay dedicated and when SR-IOV virtual functions, VMs, or sidecar isolation can safely share it.
The Good Stuff
Paid picks we use ourselves. Links carry a referral code — a subscription through one earns us a commission at no extra cost to you.
Newest first
Brazilian payment processor Trio faced severe data challenges after scaling and acquiring PayBrokers, processing over 243 million payments in early 2025 across disparate legacy databases. Because financial reconciliation and compliance demand absolute precision, the engineering team unified their analytics layer by migrating to ClickHouse Cloud. To handle out-of-order and duplicate events from their Redpanda streaming pipelines, Trio implemented sliding-window refreshable materialized views alongside ReplacingMergeTree engines rather than recomputing entire historical datasets. A custom ETL service manages schema drift before inserting clean records into ClickHouse, while batch ingestion is coordinated through Airflow. Consequently, Trio reduced storage consumption by approximately 88% through columnar compression, ingested a 5-billion-row migration without issue, and maintained real-time query performance across more than one billion daily ledger data points.
ClickHouse 26.3 LTS introduces experimental materialized common table expressions, WebAssembly user-defined functions, and enables asynchronous inserts by default. Alongside the release, Alasdair Brown launched clickhousectl to manage local installations and cloud deployments with built-in agentic development features and authentication safeguards. User case studies show Goldsky accelerating historical blockchain backfills by twelve times through direct ClickHouse reads using Apache Arrow, replacing their prior Kafka and Avro architecture. Additionally, the engineering team redesigned the full-text search index with front-coding compression, a sparse lookup index, and adaptive posting lists to handle expensive random reads on object storage. ClickHouse also previewed community architecture patterns for lightweight SIEM setups using Clickdetect and Wazuh log collection.
Zodiac, Palantir’s custom 3D globe library built on Three.js, encountered severe performance degradation near the Arctic because latitude/longitude tiles converge into tiny slivers and create excessive geometry. Standard equal-angular tiling was replaced internally with polar-scaled tiles whose widths expand toward the poles so their surface areas remain closer to equatorial tiles. The design uses spherical surface-area calculations, power-of-two merge spans, and dynamic quadtree expansion; client-side texture compositing handles standard geodetic or Web Mercator sources, while heightmaps are combined and resampled for wider tiles. The change reduced polar geometry, draw calls, and textures by over 90% and made frame time more consistent, though texture loading remains the next bottleneck and native polar tiles would remove stitching overhead.
VAKRA is an executable, tool-grounded benchmark for evaluating whether AI agents can complete enterprise-like workflows that combine API calls, document retrieval, and natural-language constraints. It provides more than 8,000 locally hosted APIs backed by real databases across 62 domains, plus domain-aligned documents, and uses full execution traces to assess multi-step behavior. The four capability groups cover API chaining, dashboard-API tool selection, multi-hop dashboard reasoning, and multi-hop multi-source reasoning with multi-turn dialogue and optional tool-use policies; tasks can require three to seven reasoning steps, while some segments specify one to twelve tool calls or up to five logical hops. Reported analyses find that models perform best on single-hop questions, degrade as API hops increase, and often violate policies or fail to retrieve enough information when constraints restrict the relevant source, exposing a gap between isolated tool competence and end-to-end reliability.
Repeated visual-regression failures made Polaris CI feedback cycles take about 30 minutes, prompting an attempt to optimize build time rather than fix individual tasks. The approach uses Pi to create an Autoresearch extension that repeatedly forms a hypothesis, measures the build against a 19.1-second baseline, and keeps only changes that improve the metric, discarding crashes and slower runs. The loop identified wasted VRT work: the full component pipeline ran before Storybook recompilation, while TypeScript transformed all 580 component files although 105 needed it. Removing those inefficiencies made the build 65% faster, although the system also produced unacceptable hacks that were discarded. The extension later gained multi-metric support, consistent iteration scripts, auto commits, and open-source adoption, with reported internal improvements including unit tests running 300 times faster and Playwright tests becoming faster.
ClickHouse accelerates analytical queries by minimizing the amount of data read from disk across granules. A primary key determines data part sort order and records the first value per granule, enabling binary search to skip unneeded granules during filtered queries. For non-primary key filters, lightweight projections act like secondary indexes by storing alternative sort keys alongside base table offset pointers without duplicating full table rows. Skip indexes, such as minmax indexes, record column boundaries per granule and operate without disk duplication, though effective pruning requires some correlation with the primary key. Demonstrations using a UK property price dataset show how primary key filters and minmax skip indexes eliminate large fractions of total granules.
Documentation platform Mintlify experienced major dashboard performance degradation and rate limiting under PostHog as multi-tenant analytics traffic grew rapidly. Because PostHog lacked incremental materialized views, queries scanned full raw datasets and took tens of seconds to load. Mintlify migrated its customer-facing analytics infrastructure to ClickHouse Cloud, ingesting events through Amazon MSK and ClickPipes into a single events table backing incremental materialized views. The team executed the transition over two weeks by dual-writing events and replicating the existing PostHog schema to preserve API surface compatibility. The migration reduced dashboard query latency to sub-one-second, decreased related support bug reports to zero, lifted NPS by an estimated 30%, and lowered analytics infrastructure costs by approximately 60%.
The post reports a three-week experiment in which a multi-agent system optimized 235 CUDA kernel problems generated from workloads spanning more than 124 production open-source models. Using SOL-ExecBench on 27 NVIDIA Blackwell 200 GPUs, the system used a planner to distribute and rebalance work, while autonomous workers benchmarked, debugged, and refined kernels without developer intervention. It ran separate CUDA C with inline PTX and CuTe DSL runs, reaching assembly-level optimization and learning a documented higher-level API. Against a single-agent optimized PyTorch baseline, it outperformed 149 of 235 problems, achieving a 1.38x geometric mean ratio, while 45 solutions exceeded 2x. The highlighted cases included an 84% geomean gain over a human FlashInfer baseline for paged-prefill attention, a 39% gain for NVFP4 MoE linear, and a GEMM reaching 86% of a tuned cuBLAS baseline; median SOL was 0.56, leaving room for improvement.
Ramp built automatic receipt detection for its iOS app to reduce the effort of finding receipt photos while keeping private images on device. An initial design used PhotoKit, Vision OCR, background processing, local storage, and exact checks for merchant, amount, and date, but formatting differences, aliases, and OCR errors produced many false negatives. FoundationModels enabled a local structured-output LLM, yet a single prompt performed poorly in evals because the small 4096-token model hallucinated and took shortcuts. Splitting the merchant, date, and amount questions and requesting rationale improved evaluation results, but the alpha release reached 66% precision and about 18% recall. Version 3.0 combined one LLM merchant-name check with deterministic iOS 26 data detectors for dates and monetary values; it cut processing time, more than doubled recall, and raised precision to 87%.
Overworld’s Waypoint-1.5 is a real-time video world model aimed at making interactive generative worlds usable on consumer hardware rather than limited to datacenter-scale compute. It offers two tiers: a 720p model for higher-performance desktop GPUs, capable of up to 60 FPS on RTX 3090 through 5090 hardware, and a 360p model for broader deployment, including gaming laptops and soon Apple Silicon Macs. Training used nearly 100 times more data than Waypoint-1, while more efficient video-modeling techniques reduce redundant computation across frames to support coherent motion and responsive exploration. Users can run it locally through the Biome desktop client, try it in a browser via Overworld Stream, or use the World Engine inference library and third-party clients. The release presents improved visual fidelity, broader local hardware support, and continued progress toward real-time interactive world simulation.
Stripe’s Ruby monorepo contains about 50 million lines, 100,000 test files, and roughly 1.2 million test units, making sequential execution impractical at an estimated four months. To keep CI within minutes, Selective Test Execution (STE) runs tests whose recorded file accesses intersect files changed since a suitable baseline, averaging about 5% of the suite and under 10% of always-run compute. A dynamically linked C++ library loaded with LD_PRELOAD intercepts file opens, associates them with hierarchical test scopes, and propagates tracing to child processes; logs are aggregated into a selection index outside the syscall hot path. Guardrails cover tests affected by directory discovery, previously failing tests, selective linting, generated files, and reproducible database-backed baselines ordered by Monotonic Revision IDs. The system is presented as a way to preserve safety and affordability while scaling CI for a highly dynamic language.
The ClickHouse community and development team participated in FOSDEM 2026 in Brussels, hosting a community dinner and delivering technical presentations across multiple developer tracks. Presentations covered a range of engineering topics, including building geospatial analytical applications with Leaflet and ClickHouse, as well as integrating LLVM XRay to dynamically hot-patch production logging without redeployments. Additional ClickHouse sessions addressed the engineering challenges of integrating Rust libraries into large C++ codebases, optimizing low-level memcpy routines, and implementing specialized inverted indexes for large-scale text processing. Contributed community sessions detailed workflows for upstream patches in MariaDB and Postgres, alongside WebShield, an open-source Python daemon that streams web accelerator access logs into ClickHouse for bot analysis.
Safetensors has joined the PyTorch Foundation as a foundation-hosted project under the Linux Foundation, alongside DeepSpeed, Helion, Ray, vLLM, and PyTorch. It began as a Hugging Face project addressing the risk of arbitrary code execution in pickle-based model-weight formats. Its format stores tensor metadata in a JSON header capped at 100MB, followed by raw tensor data, and supports zero-copy mapping and lazy loading. The move places the trademark, repository, and governance with the Linux Foundation, while Luc and Daniel remain on the Technical Steering Committee and lead day-to-day work. Existing users get the same format, APIs, and Hub integration without breaking changes, while contributors gain a documented maintainer path; planned work includes PyTorch core serialization, accelerator-aware loading, parallel loading, and quantization support.
Eisan System Development moved a retail analytics workload from a QlikView-based platform to ClickHouse Cloud after rising ID-POS data volumes caused slower processing, failed analyses, and increasing licensing and server costs. Client records from supermarkets and drugstores support basket analysis, cross analysis, and real-time personalization, with some datasets nearing 200 million records annually. The team compared open-source options including PostgreSQL and YugabyteDB, then tested ClickHouse configurations, including a 30-core, 120-GB setup, while accounting for cache memory usage through load testing. In comparison, ClickHouse ran on slightly lower-spec hardware under high-load conditions without observed performance deterioration, and Eisan reports no cases so far where it failed to return a result; the team is now targeting 100 billion records and exploring off-hours processing and scheduled scaling.
MongoDB Atlas initially relied on a reactive auto-scaler that took minutes to adjust instance sizes and only transitioned between adjacent tiers. Because these delays caused transient server over-utilization or unnecessary customer costs, researchers prototyped predictive auto-scaling to forecast demand. The experimental architecture used short-term and long-term forecasters for customer-driven metrics, an estimator using boosted decision trees to project CPU utilization, and a planner to pick the most cost-effective tier. In testing, the prototype kept replica sets closer to target CPU ranges while saving an average of nine cents per replica set hourly. MongoDB subsequently implemented a production version that scales replica sets up proactively before expected spikes while retaining reactive downscaling.
Palantir’s frontend engineers are building a 3D implementation of Gaia to visualize Foundry Ontology objects with geospatial properties, including ground locations, satellite positions, and paths through space. The system uses Zodiac, a Three.js library backed by Foundry-based APIs: the backend stores, filters, and computes data, while the frontend receives simplified geometry and style information for rendering. Performance techniques include converting latitude and longitude to Earth-fixed coordinates once, batching similarly styled lines into single draw calls, and using instanced meshes plus batched updates for dense satellite scenes; orbital propagation uses TLEs client-side, while heavier analysis remains server-side. The article also describes frame-synchronized animation callbacks that update material properties without rebuilding geometry, while noting that 3D APIs continue to iterate across usability, platform integration, and performance as new data sources and workflows are added.
ClickHouse 26.3 delivers twenty-seven new features, forty performance optimizations, and over two hundred bug fixes. Common table expressions can now be materialized into temporary tables via the MATERIALIZED clause when enabled, evaluating subqueries once to reduce query execution times and peak memory usage. The query optimizer expands join reordering beyond inner and left or right joins to include anti, semi, and full joins when table statistics are available. In addition, user-defined functions can run compiled WebAssembly code inside sandboxed Wasmtime runtimes. Storage for the Map data type also introduces hash-based bucketing during background merges, accelerating single-key lookups by two to forty-nine times without slowing initial inserts.
Yelp's Database Reliability Engineering team upgraded over a thousand Cassandra nodes from version 3.11 to 4.1 without downtime. The upgrade delivered reliability, observability, and performance improvements, including Java 11 support, configurable guardrails, and fixed incremental repairs. To avoid the high costs, consistency compromises, and data streaming delays of standing up separate datacenters, the team selected an in-place rolling upgrade strategy. The rollout orchestrated staged node updates alongside compatible versions of the Stargate proxy and custom data pipeline connectors without requiring client code modifications. Although Stargate 2.x introduced regressions on range and multi-partition queries that required downgrading to 1.x, the finished migration achieved up to a 58% reduction in p99 latency across key clusters.
MoE inference commonly organizes decode around experts, but the post describes “warp decode,” an output-centric kernel for small-batch generation on NVIDIA Blackwell GPUs. Instead of staging tokens into expert-major buffers, each warp owns one output scalar, streams the required weight rows, accumulates contributions from all eight routed experts in FP32, and uses warp-level reductions; the gate/up and down projections are compressed into two kernels. This removes padding, scatter, combine, activation gather, and per-expert output buffers, while keeping BF16 activations and avoiding intermediate quantization. On B200s, the approach delivered 1.84x higher decode throughput, sustained 3.95 TB/s at batch size 32, and produced outputs 1.4x closer to a full FP32 reference.
After introducing a direct erasure-coding service, Dropbox encountered severe storage fragmentation in its immutable blob store, Magic Pocket, due to a large number of under-filled volumes. Because their legacy compaction strategy was only designed to top off near-full volumes, the team created new multi-strategy compaction methods to aggressively consolidate sparse volumes. These approaches leverage dynamic programming and streaming pipelines to rapidly reclaim unused disk capacity at exabyte scale.
The best engineering reads, once a week.
