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
CRuby’s original garbage collector stores most Ruby objects in fixed 40-byte RVALUE slots, leaving differently sized data partly unused or forcing additional allocations through malloc. Shopify’s Variable Width Allocation project adds dynamic-sized allocation APIs and five fixed-size pools—40, 80, 160, 320, and 640 bytes—while increasing heap pages from 16 KB to 64 KB. The design targets cache locality, malloc overhead, memory utilization, and future garbage-collector experimentation; a new GC.stat_heap API reports size-pool statistics, and an escape hatch can disable the feature at build time on Ruby 3.2. Benchmarks on development Ruby 3.2.0 showed gains in most workloads, including 9.8% for psych-load, 5.3% for hexapdf, and 2.1% for railsbench, while mail was 1.7% slower.
Aaron Patterson’s 2022 review surveys developments in Ruby and Rails, pairing runtime performance work with tools intended to improve developer productivity. YJIT’s compiler was rewritten in Rust, enabling ARM and Apple Silicon support while producing faster machine code; Variable Width Allocation adjusts object slot sizes to keep instance variables embedded and improve data locality. Ruby 3.2’s Object Shapes use property order as a class-independent inline-cache key, while Ruby LSP adds editor services and Error Highlight adds column-level diagnostics that now appear on Rails error pages. Patterson predicts—more accurately, hopes—that languages, frameworks, and development environments will become more interconnected, providing better suggestions, context, and assistance so developers can spend less time coding and more time solving problems.
PostgreSQL 15, released as a stable version in October 2022, adds security, performance, SQL, replication, logging, compression, and string-processing changes. CREATE permission is revoked from all users except the database owner, views can use caller permissions with security_invoker=on, and sorting is reported up to four times faster for in-memory and on-disk workloads; window functions also improve. The release introduces MERGE, supports two-phase commits and selective data in logical replication, allows conflicting transactions to be skipped or subscriptions disabled, and enables parallel workers for SELECT DISTINCT. It also adds JSON logging, Zstandard and ZL4 compression for WAL and pg_basebackup, and POSIX regexp functions such as regexp_substr() and regexp_count(). The post recommends upgrading for bug fixes and stability, while noting that migrated databases may need manual permission changes and upgrades should be checked for rare manual actions.
pg_graphql 1.0 marks the general availability of a PostgreSQL extension that lets databases be queried with GraphQL and formalizes a SemVer-based guarantee against backward-incompatible changes. It was designed for Supabase Free Plan servers with 1 GB of memory, where alternatives such as Hasura and Graphile consumed more memory than the platform could sacrifice. The implementation began with a C parser and SQL transpiler exposed through PostgREST RPC, then moved from SQL to Rust after nested SQL functions, copied state, and JSONB AST processing slowed schema introspection; v0.5.0 completed the rewrite. On Free Plan hardware, version 1.0 adds less than 300 microseconds per request, reaches about 1,060 SQL queries per second per connection, and the full Auth and HTTP stack handles about 645 requests per second, while larger instances exceed 10,000 requests per second.
Shopify has adopted Rust as its official systems programming language while retaining Ruby for server-side business logic, and has joined the Rust Foundation. The change addresses growing systems-programming needs including high-performance network servers, Ruby native extensions, and WebAssembly, where Shopify seeks speed, productivity, safety, and community-driven open source. Shopify cites Rust’s predictable native-code performance, fine-grained memory control, compile-time safety, support for concurrency, and interoperability with C and Ruby through tools and crates such as bindgen, rb-sys, and magnus. The company says Rust performed strongly in its safety evaluation and that its initial projects exposed more errors at compile time, while acknowledging that performance still depends on design and measurement and that internal education, tooling, and community participation remain unfinished work.
Supabase Storage introduces image resizing, webhooks, and a Smart CDN, positioning them as components of an on-demand image delivery system. Resizing uses imgproxy and supports width, height, and contain, cover, or fill modes through JavaScript clients or direct Storage API requests; images are not enlarged beyond their original dimensions, and aspect ratio is preserved by default. The Smart CDN stores resized results using cache keys based on an Etag, tenant ID, bucket, and image name, while webhook-driven metadata updates propagate invalidations globally within about 60 seconds. Image Resizing is beta for Pro and Enterprise customers, with 100 transformed images per project at no additional cost and $5 per additional 1,000; self-hosting is free.
Shopify presents three additional practices for optimizing large stateful Apache Flink applications, focusing on parallelism, sink capacity, and combining heterogeneous sources. It recommends starting with execution-environment parallelism, then matching task-manager capacity to the highest parallelism value; for example, parallelism 100 requires 25 task managers with four slots each. Sink bottlenecks can propagate backpressure upstream, so batching writes and correcting data skew—including with key bucketing when a low-cardinality or uneven key is unavoidable—can improve distribution, though bucketed results must later be combined. For archived Kafka topics, HybridSource exposes cloud-storage history and the live Kafka topic as one ordered logical DataStream, automatically switching after the archive is exhausted and using object-storage partitions to accelerate historical backfills.
Supabase announces pg_crdt, an experimental Postgres extension that adds CRDT support as a data type for collaborative and offline-oriented applications. It frames CRDTs as structures whose updates can be merged in any order to reach the same final state, avoiding the last-write-wins behavior of ordinary arrays. A content column can use crdt.ydoc, backed by Yjs, while the extension also supports Automerge through Rust libraries wrapped with pgx. The post contrasts storing CRDTs directly in Postgres with making Realtime an authoritative server, favoring the database-centered approach for simpler architecture and broader tooling, while noting that the decision is unsettled and pg_crdt is not released on the Supabase platform. Known limitations include full-document WAL broadcasts, increased WAL and dead tuples for frequently updated documents, and serialization overhead for large CRDT values.
Supabase describes the hiring traits and operating model it believes support its growth, arguing that people matter more than methodology. It looks for extreme asynchronous effectiveness, egoless execution, continuous incremental improvement, a bias toward immediate action, ambitious problem-solving, and a founder mindset; seniority does not exempt anyone from independent-contributor work, self-management, or maintenance. The 60-person team spans 25 countries, uses meetings as a last resort, and gives autonomous product teams access to shared metrics, user feedback, and equity through an ESOP. Teams are organised around products such as Postgres, Auth, Storage, Realtime, and Functions, while cross-cutting groups consult rather than merely receive delegated tasks. Three-month product cycles set ambitious goals, and Kaizen Weeks focused on QA, documentation, and performance help address technical debt and incorporate improvements into ongoing workflows.
Grab's Geo team required an efficient, low-cost way to collect fresh street-level imagery across Southeast Asia, where professional mapping equipment is prohibitively expensive and smartphone crowdsourcing yields inconsistent quality. To address this, Grab built KartaCam, a custom mapping device equipped with a 12MP sensor, dual-band GNSS, 4G LTE, and onboard edge AI. Edge machine learning models evaluate scene suitability, check image quality, filter for map-relevant objects, and blur personal data directly on the device prior to upload. Deploying a four-camera KartaCam 360 array delivers panoramic coverage and point-of-interest data comparable to commercial mapping rigs at roughly one-twentieth the hardware cost.
Shopify’s 2022 BFCM Live Map needed to deliver real-time sales and product data to thousands of concurrent users while processing data from millions of merchants. Its 2021 architecture used WebSocket delivery to a presentation layer, which stored messages in a mailbox that clients polled at least every 10 seconds, creating latency and bottlenecks. Shopify replaced that path with a streamlined Flink-based pipeline and a Golang SSE server that subscribes to Kafka topics and pushes JSON data to registered, authenticated clients as soon as it is available. The SSE server ran horizontally behind Nginx load balancers, with a configurable Java client used to load-test concurrent connections. The 2022 system maintained 100 percent uptime, delivered SSE data within milliseconds of availability, and visualized data within 21 seconds of creation, including pipeline processing.
The post compares SQL and NoSQL storage for an application that records daily food, water, exercise, weight, and notes, focusing on the tension between easy document writes and later analysis. It proposes using PostgreSQL as a hybrid: keep core fields such as date, user_id, weight, and notes in relational columns, while storing food_log, water_log, and exercise_log as JSONB fields. For analysis, PostgreSQL's jsonb_array_elements expands arrays into rows; ->> extracts text, -> returns JSON, and numeric values such as calories must be cast to INTEGER before aggregation, with a common table expression used to work around the restriction on set-returning functions inside aggregate calls. The examples show daily calorie totals and a search for garlic bread returning 200 calories, supporting the conclusion that JSONB can simplify storage while SQL preserves querying and analysis capabilities.
Shopify needed a reliable way to measure React Native rendering times and verify that mobile apps remained as fast as native while adopting React Native. The open-source @shopify/react-native-performance library measures app startup, navigation, and screen re-render times by timing from native startup, a user action, or a UI event until a screen is fully rendered and interactive. It uses an invisible marker view with a native counterpart to capture end-to-end timing, including React Native-to-native bridge communication, and models incremental rendering as interactive or non-interactive render passes in a state machine. Reports can be sent through onReportPrepared to analytics tools such as Amplitude, where percentile TTI dashboards compare screens over time, expose bottlenecks, and help catch regressions; Shopify also uses screen-level Apdex internally.
FlashList is Shopify’s React Native list library, created after Shopify’s migration from native technology exposed poor list performance in React Native. The post explains how a four-person team prototyped a library that combined FlatList’s familiar API with RecyclerListView’s performance, while addressing blank spacing and low frame rates on some devices. It also describes a launch planned six months ahead around a landing page, documentation, blog storytelling, social media, conference exposure, and feedback from external users. The post says FlashList was used in production on some Shopify app screens, and the launch achieved 1,000 GitHub stars in less than 24 hours. The broader conclusion is that successful open-source projects require product thinking, user empathy, community building, and a long-term authority-building strategy in addition to code.
Shopify’s migration from Apollo GraphQL client 2 to client 3 prompted a closer examination of the Apollo InMemoryCache after past bugs were linked to misunderstanding or misuse. The cache stores an in-memory representation of queried data for the current browser session, and its lifecycle covers fetching, normalization, updating and merging, then garbage collection and eviction. Fetch policies determine whether data comes from the cache, the network, or both; under the default cache-first policy, incomplete data triggers a network request. Normalization breaks responses into objects, assigns cache identifiers usually from __typename and id or configured key fields, and stores them in a flattened structure. Automatic UI updates depend on matching identifiers and suitable query or mutation responses, while objects made unreachable by changed identifiers can remain until garbage collection removes them.
During the infrastructure work for a marketing tool, Shopify's team found a BigQuery query that would have processed about 75,462,743,846 bytes per request and cost nearly $1 million monthly at an estimated 60 requests per minute. The query served a pipeline ingesting one billion rows through Apache Flink, with state managed by RocksDB and streaming requests from Apache Kafka; scaling beyond the release made ingestion unsustainable. To support general availability, the team evaluated an external SQL warehouse that could load Parquet atomically, handle 60 requests per minute, and export results to Google Cloud Storage, then clustered a dataset on two feature columns used in WHERE clauses. Running the same query on the clustered table reduced billed data to 508.1 MB and identified 108.3 MB scanned, lowering estimated monthly cost to about $1,370.67; the post also recommends selecting needed columns, partitioning tables, and using free previews instead of exploratory queries.
Shopify announces that the open-source web framework Remix and its team are joining the company to help developers deliver lightning-fast, resilient web experiences. Remix combines multi-page and single-page strengths through a full-stack React framework with data loading and code splitting handled without requiring front-end developers to manage those backend tasks. It also emphasizes progressive enhancement, with links, buttons, and forms working before JavaScript loads, and supports edge, service-worker, and Node runtimes. Shopify will use Remix across projects including Hydrogen; Hydrogen is moving from server components to Remix's data-loading pattern for faster performance and a simpler developer experience, with ~90% of developers' code expected to remain unchanged. Remix will remain independent and open-source while Shopify supports its roadmap and community.
PostgreSQL Commitfest is the community's month-long review process for patches intended for merging into core. Contributors submit patches to pgsql-hackers, create a Git branch from master, consolidate changes, generate a versioned patch, and associate its mailing-list thread with an open Commitfest entry. The process expects authors to explain rationale, respond to review, rebase or revise patches, and review a comparable number of other patches. Submitted changes are checked by CFBot across architectures and operating systems, while statuses such as "Needs review," "Ready for committer," and "Waiting on author" track progress. Commitfests occur five times yearly, with the March event tied to feature-freeze for the October major release.
The post compares PostgreSQL Full Text Search with dedicated search engines including MeiliSearch, OpenSearch, SQLite FTS, and Typesense, asking whether a separate system is justified for common application search. It presents a PostgreSQL setup using a generated stored column built with to_tsvector('english', ...) and a GIN index, queried with websearch_to_tsquery. The benchmark loads a movie dataset and focuses on query speed after data ingestion, using comparable scripts and default configurations with limited tuning. PostgreSQL was close in performance to several alternatives; in-memory SQLite was fastest, while Typesense and MeiliSearch handled misspellings correctly and OpenSearch ingested data quickly. The conclusion is that PostgreSQL can provide good-enough search quality and speed for basic cases without adding another system, although dedicated engines offer capabilities PostgreSQL does not.
Supabase and Snaplet open-sourced postgres-wasm, a PostgreSQL server that runs inside a browser and supports browser persistence, pg_dump restoration, and logical replication with a remote database. Rather than compiling PostgreSQL directly to WebAssembly, the project packages a stripped-down Linux image with PostgreSQL in a Buildroot-based virtual machine, then runs it through v86 in the browser. PostgreSQL 14 startup failures were resolved by disabling v86 JIT compilation, while image optimization used a compressed 9P filesystem, asynchronous HTTPS loading, and page_poison=on to produce an initial state of about 12MB. Websocket proxying provides outbound connectivity and a forked proxy with nginx tunnels incoming connections to PostgreSQL. The project remains experimental because the download is currently about 30MB, but the post identifies offline data, testing, development environments, snapshots, and support as potential uses.
The best engineering reads, once a week.
