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
The Shop app’s Add to Favorite interaction animates a product thumbnail into the Favorites tab bar heart after a buyer favorites a product. The implementation is based on a motion spec that defines animated properties, target values, durations, easing curves, and ordering, with the thumbnail sequence followed by the icon’s color and bouncing movement. In React Native, Reanimated uses one shared progress value for the whole sequence, and interpolation maps progress ranges to opacity, scale, and vertical translation, including clamping and an instant disappearance at step 3. A 50 ms pause is included between thumbnail scaling and its downward movement. The shared value also creates a match cut between the disappearing thumbnail and the icon; a setTimeout-based handoff was rejected because a busy JS thread could delay the transition, while withDelay would keep timing on the UI thread.
Production deployments of BERT and Transformer architectures often face cost and latency challenges because these models are significantly larger and more computationally intensive than traditional algorithms. To optimize text classification workloads, developers can compile vanilla PyTorch models for AWS Inferentia using the AWS Neuron SDK and its tracing utilities. Because the Neuron SDK requires static tensor dimensions, the model is traced with fixed input lengths and packaged alongside a custom inference script configuring one Neuron Core per worker. Deploying the resulting artifacts to an Amazon SageMaker ml.inf1.xlarge endpoint yields an average latency of 5 to 6 milliseconds for a sequence length of 128 across 10,000 synchronous evaluation requests.
Liquid-C is an extension to the Liquid gem that parses, tokenizes, and renders templates, but some Liquid features, including If and For tags, still render in Liquid rather than Liquid-C. The work explains how Liquid-C compiles templates into bytecode for a customer VM and addresses the difficulty of moving instruction and constant pointers when control flow requires forward or backward jumps. It refactors instructions to store each constant’s index in the constants array, using a hash table to avoid duplicate constants and remove the constant pointer. The change reduced one example constants array from 157 objects to 104, but did not drastically improve theme rendering speed; it slightly increased memory usage and parsing time while preparing for GOTO or JMP operations and later If and For support.
Traditional PostgreSQL auditing methods duplicate source table structures, creating maintenance overhead when schemas change. To solve this, a compact SQL implementation stores insert, update, and delete events across multiple tables in a single audit table using JSONB columns. Query performance is maintained by using a BRIN index on naturally ordered insertion timestamps, a B-tree index on internal table OIDs, and UUIDv5 identifiers hashed from primary keys. Row-level PL/pgSQL triggers automatically populate the audit log and can be enabled or disabled dynamically with dedicated tracking functions. While trigger-based auditing introduces minimal overhead at rates below 1,000 writes per second, write-heavy workloads may benefit from logging changes outside SQL with tools like pgAudit.
Shopify presents lessons from operating large, stateful Apache Flink applications on Kubernetes with Google Kubernetes Engine, RocksDB, and checkpoints and savepoints stored in Google Cloud Storage. The guidance begins with profiling tools, including async-profiler, VisualVM, jemalloc with jeprof, and Eclipse Memory Analyzer, to investigate CPU, heap, native-memory, and leak-related problems. Disabling Kryo fallback exposed unsupported serializers such as Scala BigDecimal and Scala algebraic data types; replacing them with supported alternatives increased throughput by 20 percent. The post also recommends separate configuration profiles for high-throughput backfills and low-latency steady state, while describing a RocksDB native-memory failure in which disabling the block cache stopped repeated out-of-memory crashes without affecting application performance.
Shopify’s core monolith contains more than 170,000 tests, and variable CI feedback times create delays and context switches as the test pool grows. The Test Infrastructure team evaluated a Test Budget approach that builds prioritized orders on top of deterministic test selection, then stops execution at a predetermined time to increase the chance of finding failures early while accepting some risk. Six criteria were compared—failure_rate, avg_duration, churn, coverage, complexity, and default/random—using Time to First Failure, APFD, and Convergence Index; historical results were stored and queried with Redis Sorted Sets. The analysis found failure_rate had the best detection rate, while churn and complexity were inappropriate; running 60% of the suite detected 80% of failures in the mean case, and 70% of the test-selection suite detected 50% at the fifth percentile.
v8go adds support for V8’s CPU Profiler, allowing Go users to measure JavaScript execution in a V8 isolate and inspect sampled top-down call trees with function and engine details. The API work aimed to stay idiomatic for Go, remain close to V8’s API, support future profiler features, and minimize Cgo overhead. The first design loaded profile data lazily, making separate Cgo calls for profile and node properties; the alternative built the complete graph in C++ before returning it to Go. Benchmarks showed lazy loading was faster for a small tree, averaging about 20 microseconds versus 25 for eager loading, while a larger tree reversed the result at roughly 90 versus 60 microseconds. The comparison demonstrates that Cgo costs depend on workload size and that performance assumptions require benchmarks using representative inputs.
RubyConf 2021: The Talks You Might Have Missed compiles talks presented by Shopify’s Ruby and Rails infrastructure engineers at the November conference in Denver, offering ways to revisit sessions for attendees who missed them. The program covers Ruby implementation topics including compiler history, Aaron Patterson’s pure-Ruby JIT compiler, YJIT’s incremental integration inside CRuby, parser generators, variable-width allocation, and Ruby memory layout. Other sessions address gradual Sorbet adoption, native extension compilation and security, pair programming, memoization performance, executable code as data, Ruby archaeology, fast CI, and Ractor-based parallel testing. The roundup reports early YJIT performance results, describes a CI environment with 2.8 million Rails monolith lines and 210,000 Ruby tests, and notes variable-width allocation passed Shopify’s Rails CI while serving more than 500 million requests in a week of limited production traffic.
React Server Components (RSC) introduced a paradigm shift for the team building Hydrogen, a React-based framework for custom storefronts. The post presents patterns developed through trial and error, beginning with shared components as a middle ground and reserving client components for interactivity, useState or useReducer, lifecycle logic, unsupported libraries, or browser APIs. Components should instead become server components when they contain secrets or proprietary logic, need filesystem or database access, fetch storefront API data, or do not execute on the client. The examples extract newsletter form controls and FAQ accordion behavior into client components, then rename the remaining wrappers as server components when their usage allows it; the FAQ server component is passed through a client component as children.
Updated for Hydrogen 0.26.0, the tutorial builds a simple product page for a custom Shopify storefront using Hydrogen’s React-based framework. Starting from the Hello World template on StackBlitz, it adds Tailwind CSS, creates a file-based dynamic product route, and queries product data through the Storefront API with the useShopQuery hook. The implementation then uses React Server Components alongside a client component and ProductOptionsProvider to track selected variants and options, update prices, and support a BuyNowButton checkout flow. A media gallery and additional styling complete the page, which uses products and media from the Hydrogen Preview store. The post concludes that Hydrogen’s components, hooks, and utilities reduce repetitive storefront work and speed development.
Shopify’s 2021 BFCM four-day event exposed the scale its commerce platform must absorb: peak app-server traffic exceeded 32 million requests per minute, load balancers surpassed 34 million RPM, and a flash sale used more than 20% of total computing capacity. The post frames larger, spikier sales, mobile and social commerce, and expanding buying environments as enduring engineering challenges for supporting millions of merchants. Planned work includes evolving the large Rails codebase into modular components, moving storefront rendering out of the monolith, expanding Spin cloud development environments, and making in-house load testing part of product development. The storefront split and refactoring produced four-times-faster cache fill rates and five-times-faster page rendering, while YJIT ran 31% faster than interpreted CRuby and 26% faster than MJIT; Shopify also remains committed to Rails and React Native.
Deploying large Transformer models at scale often faces severe prediction latency bottlenecks, driving up infrastructure costs and limiting real-time production use cases. To address this challenge, Hugging Face evaluated Infinity, a containerized hardware-optimized inference solution paired with the Infinity Multiverse model optimization service. Testing covered 192 configurations on Amazon EC2 C6i instances powered by 3rd generation Intel Xeon Scalable processors across varying CPU cores, sequence lengths, and batch sizes. The benchmarks demonstrated that an Ice Lake-optimized DistilBERT container achieved up to 800% higher throughput than vanilla Transformers and delivered 1 to 4 millisecond end-to-end latencies for sequence lengths up to 64 tokens. Although Infinity was later discontinued in favor of Inference Endpoints and Optimum libraries, the results demonstrated substantial efficiency gains on modern CPU hardware.
Wav2Vec2 models fine-tuned with Connectionist Temporal Classification transcribe speech without external language models, but decoding can still suffer from spelling inaccuracies. Hugging Face Transformers addressed this by integrating Kensho Technologies' pyctcdecode library to support decoding with n-gram language models. Instead of decoding using simple argmax operations over logits, the Wav2Vec2ProcessorWithLM class feeds full probability matrices into beam search guided by KenLM n-gram probabilities. KenLM's build_binary utility compresses language model ARPA files into binary formats, reducing file size by more than half for faster loading and hub deployment. In Swedish xls-r-300m-sv benchmarks on Common Voice 7, this 5-gram boosted decoding setup achieved an 18.85% word error rate, delivering an approximate 30% relative performance gain.
Shopify’s year-end roundup curates ten favorite engineering posts from 2021, spanning Ruby, storefronts, data analysis, databases, search, GraphQL, CI, APIs, and mobile development, while also noting several older reader favorites. The selection includes tutorials on building a web app with standard Ruby libraries, exploratory data analysis, GraphQL, Apache Beam for search, and React Native App Clips, alongside technical accounts of Hydrogen, YJIT, MySQL upgrades, and GraphQL rate limiting. YJIT was merged upstream and included in Ruby 3.1.0, while the CI team reduced the p95 of Shopify’s core monolith from 45 minutes to 18 minutes. The roundup also describes query-cost-based GraphQL rate limiting, rollback-testing roadblocks during a MySQL upgrade, and an App Clip effort that used a fail-fast, iterative approach.
Standard Transformer architectures scale poorly in compute and memory because pairwise dot-product self-attention depends quadratically on input size. Perceiver IO addresses this constraint by computing self-attention across a small set of latent variables rather than directly on high-dimensional inputs. Inputs and outputs interact with the model via cross-attention operations, decoupling compute and memory costs from input and output dimensions. Integrated into Hugging Face Transformers via the PerceiverModel class, the architecture supports diverse data types using optional preprocessors, decoders, and postprocessors. Experiments demonstrate competitive performance across text, multimodal video classification, 3D point cloud classification on ModelNet40, and StarCraft II reinforcement learning in AlphaStar.
Shopify’s 2021 BFCM live map had to scale from more than 1.7 million merchants while adding product trends and unique shoppers, after the prior Cricket-and-Redis backend showed volume and delivery weaknesses. The redesign placed Apache Flink on the critical path to filter irrelevant checkout events, reducing Cricket’s workload to about one percent of event volume, while cross-region sharding and active-active deployment supported availability. Flink computed new metrics, including 500 product categories ranked by sales-volume change every five minutes, while Cricket retained existing metric computation and deduplication. In load tests, the pipeline handled 1 million checkout events per second, and its jobs ran at 100 percent uptime during BFCM without backpressure or manual intervention. Redis still caused high CPU and memory pressure, producing brief arc-visual errors for a small sample of users, so Shopify plans a simpler Flink-centered design.
Supabase announced a series of platform updates, technical releases, and community initiatives to conclude its Launch Week. The team open-sourced pg_graphql, an in-progress PostgreSQL extension that keeps schema generation, query parsing, and resolvers natively on the database server. Supabase also enabled a global CDN spanning over 250 cities, delivering edge caching with a default one-hour cache time for storage assets alongside TLS 1.3, HTTP/3 over QUIC, and Brotli compression. While Supabase Functions is not yet released, two public RFCs were opened to explore architecture options, leaning heavily toward container-based execution. Finally, the update introduced a free SaaS development course with egghead.io and launched the ten-day Holiday Hackdays hackathon.
Supabase has open-sourced pg_graphql, a native PostgreSQL extension written in C and SQL that enables in-database GraphQL schema reflection and query resolution. Because free-tier virtual machines lacked memory headroom for standalone GraphQL processes like Hasura or Graphile, the team designed an in-engine solution to eliminate extra runtime overhead. The extension parses, validates, and transpiles incoming GraphQL operations into single SQL statements that aggregate data into JSON responses. This architecture avoids N+1 query problems, enforces existing row-level security policies, and maps GraphQL requests directly onto PostgreSQL transactions. The public SQL function can be queried directly or exposed over HTTP via PostgREST remote procedure calls.
Supabase updated its Realtime server to enforce PostgreSQL Row Level Security (RLS) policies when broadcasting database changes over websockets. Previously, Realtime operated as an opt-in beta feature that sent all replication changes to every client regardless of user authorization. To enforce RLS per subscriber without heavy performance overhead, Supabase introduced WALRUS, a security engine colocated inside PostgreSQL. For each replication change, WALRUS looks up active subscribers, assumes their identities, and evaluates row visibility using prepared statements queried by primary key. This in-database evaluation avoids external network round trips and single-query planning overhead while returning an authorized subscriber list to Realtime.
Transformer prediction speed can hinder latency-sensitive workloads like conversational systems and search, while optimizing hardware performance requires significant effort. Hugging Face introduced Optimum, an open-source library that reduces Transformer model latency across diverse hardware targets. Through a partnership with Graphcore, BERT is introduced as the initial IPU-optimized model for parallel AI hardware. Developers configure the Poplar SDK environment and PopTorch inside Graphcloud, install optimum-graphcore, and execute the run_qa.py script with an IPU configuration file. Fine-tuning bert-base-uncased on SQuAD v1.1 using the IPUTrainer class yields an evaluation F1 score of 88.2757 and an exact match score of 80.6623 across 10,784 evaluation samples.
The best engineering reads, once a week.
