[{"data":1,"prerenderedAt":461},["ShallowReactive",2],{"site-header":3,"site-footer-license":37,"article-rate-limiting-algorithms-token-bucket-leaky-bucket-sliding-window-with-latest":41},{"navigations":4},[5,20],{"collection":6,"item":7},"series",{"__typename":6,"title":8,"slug":9,"children_series":10},"Artikel","all",[11,14,17],{"title":12,"slug":13},"Algoritma dan Struktur Data","algoritma-struktur-data",{"title":15,"slug":16},"Konsep Dasar Pemrograman","konsep-dasar-pemrograman",{"title":18,"slug":19},"Kecerdasan Artifisial","kecerdasan-artifisial",{"collection":6,"item":21},{"__typename":6,"title":22,"slug":23,"children_series":24},"Tutorial","tutorial",[25,28,31,34],{"title":26,"slug":27},"Belajar Java","belajar-java",{"title":29,"slug":30},"Belajar Pascal","belajar-pascal",{"title":32,"slug":33},"Belajar Python","belajar-python",{"title":35,"slug":36},"Belajar Golang","belajar-golang",{"name":38,"url":39,"show":40},"CC BY-SA","https://creativecommons.org/licenses/by-sa/4.0/",false,{"article":42,"renderedBody":413,"renderedTakeaway":414,"latestArticles":415,"toc":438},{"id":43,"title":44,"slug":45,"excerpt":46,"published_at":47,"thumbnail_image":48,"tags":50,"reading_time":58,"takeway":59,"body":60,"authors":61,"series":67},"45","Rate Limiting Algorithms: Token Bucket vs Leaky Bucket vs Sliding Window","rate-limiting-algorithms-token-bucket-leaky-bucket-sliding-window","Udah tahu kenapa *rate limiting* itu wajib, tapi masih bingung beda Token Bucket, Leaky Bucket, Fixed Window, dan Sliding Window? Yuk, kita adu empat algoritma ini bareng-bareng dan cari tahu mana yang paling pas buat API kalian!","2026-09-09T01:57:55.886Z",{"id":49},"0e0ed26d-20c6-443c-bde1-4a6200f6f801",[51,52,53,54,55,56,57],"backend","system-design","rate-limiting","redis","api-gateway","architecture","traffict-management",null,"- Token Bucket paling toleran terhadap lonjakan sesaat, Leaky Bucket ngeluarin *request* dengan laju rata, Fixed Window paling gampang tapi rawan *boundary spike*, dan Sliding Window Counter paling seimbang antara akurasi dan hemat memori.\n- Token Bucket jadi standar *de facto* di *API Gateway* modern kayak AWS API Gateway, Kong, dan NGINX karena ramah sama *burst capacity*.\n- Boundary spike di Fixed Window bisa ngelolosin *request* sampai dua kali lipat kuota tepat di pergantian interval waktu.\n- Sliding Window Counter butuh Redis Hashes dan Lua Script biar operasi cek-kuota tetap atomik dan bebas *race condition* di arsitektur *microservices*.","Di artikel sebelumnya kita udah ngobrol soal kenapa *rate limiting* itu penting banget. Nah, sekarang kita masuk ke bagian yang lebih seru: algoritmanya sendiri. 🤔\n\nNyatanya, bikin *rate limiter* itu bukan sekadar ngecek `request_count > 100`. Di balik setiap *API Gateway* modern—kayak [Kong](https://konghq.com/), [NGINX](https://nginx.org/), atau Cloudflare—ada perhitungan matematis yang nentuin gimana trafik kalian diperlakukan.\n\nAda empat algoritma yang paling sering dipakai di industri. Karakternya beda-beda, dan salah milih bisa bikin kuota kalian bocor atau pengguna asli malah kena blokir. Yuk, kita bedah satu-satu!\n\n## Token bucket, si favorit yang fleksibel\n\n**Token Bucket** jadi standar *de facto* di sebagian besar *API Gateway* modern, termasuk AWS API Gateway dan Google Cloud Armor.\n\nCara kerjanya gampang. Bayangin satu ember dengan kapasitas token terbatas, misalnya 10 token.\n\n- Tiap detik, sistem nambah token baru ke ember dengan laju konstan (misal 2 token/detik). Kalau embernya udah penuh, token barunya tumpah dan dibuang percuma.\n- Setiap *request* HTTP dari klien harus ngambil 1 token biar boleh lewat.\n- Token masih ada? Silakan diproses. Token habis? *Request*-nya langsung ditolak dengan `429 Too Many Requests`.\n\nEnaknya, algoritma ini ramah banget sama *burst*. Kalau klien diem beberapa detik, embernya ngisi penuh. Dia bisa nembak 10 *request* sekaligus dalam 1 milidetik tanpa ditolak, selama rata-rata jangka panjangnya tetap sesuai kuota.\n\n*Trade-off*-nya: kalian mesti *tuning* dua parameter sekaligus. *Bucket size* nentuin seberapa besar *burst* yang ditoleransi, *refill rate* nentuin seberapa cepat kuotanya balik. Salah angka dikit, limiternya bisa kerasa terlalu longgar atau malah terlalu galak.\n\n## Leaky bucket, yang doyan ngerapiin antrean\n\nKalau Token Bucket ngatur jumlah token, **Leaky Bucket** ngatur laju keluaran *request* (*throughput smoothing*).\n\nIbaratnya ember berlubang kecil di dasarnya.\n\n- *Request* dari klien masuk kayak air dituang dari atas, ditampung di antrean FIFO.\n- Air netes keluar dari lubang dasar dengan laju stabil, misal tepat 5 *request* per detik, seberapa pun derasnya tuangan di atas.\n- Kalau tuangannya kebanyakan sampai embernya meluap, air yang luber langsung dibuang—*request*-nya ditolak dengan `429`.\n\nKelebihannya: alirannya jadi sangat stabil (*constant rate*). Cocok banget dipasang di depan *service* yang sensitif ke beban mendadak, misal sistem pembayaran perbankan warisan (*legacy system*) yang nggak sanggup nerima lonjakan.\n\nKekurangannya keras: dia nggak toleran sama *burst* sama sekali. *Request* yang sebenarnya sah malah nambah latensi karena harus ngantre di *buffer*, padahal server kalian masih sanggup kok ngelayanin.\n\n## Fixed window counter, simpel tapi ada jebakannya\n\nIni algoritma paling gampang dan paling sering ditulis *developer* pemula.\n\nWaktu dibagi jadi jendela-jendela tetap (*fixed time window*), misal per menit: 00:00-00:01, 00:01-00:02, dan seterusnya.\n\n- Tiap *request* di interval itu nambah *counter* sebesar 1.\n- Kalau *counter*-nya lewat ambang batas (misal 100 req/menit), *request* berikutnya ditolak sampai jendelanya berganti dan *counter*-nya di-reset ke 0.\n\nMasalahnya ada di jahitan antar-jendela. Anggap kuotanya 100 req/menit:\n\n- Klien nembak 100 *request* di detik 00:59. Lolos semua.\n- Klien nembak 100 *request* lagi di detik 01:01. Lolos juga, soalnya *counter*-nya baru di-reset.\n\nHasilnya, server nerima **200 *request* cuma dalam rentang 2 detik**! Kapasitas yang kalian rencanain buat satu menit penuh, dilahap habis dalam sekejap. Inilah yang namanya *boundary spike*, dan pas lagi *flash sale*, dia bisa bikin *backend* kalian tumbang. 😅\n\n## Sliding window counter, akurat tanpa boros memori\n\nBuat nutup celah *boundary spike* tanpa harus nyimpen *log* tiap *request*—itu namanya *Sliding Window Log*, dan boros memori—industri pindah ke pendekatan **Sliding Window Counter**.\n\nIdenya: hitung estimasi *request* di jendela geser saat ini, dengan nimbrung proporsi dari jendela sebelumnya.\n\n$$\\text{Estimated Requests} = \\text{Count}_{\\text{current}} + \\left( \\text{Count}_{\\text{previous}} \\times (1 - \\text{Overlap Ratio}) \\right)$$\n\nContoh, kuota 100 req/menit:\n\n- Jendela sebelumnya (00:00-01:00) nyatet 80 *request*.\n- Sekarang jam 01:15, artinya kita baru 25% masuk ke jendela saat ini. Sisanya 75% masih nempel ke jendela sebelumnya.\n- Di jendela saat ini udah ada 30 *request* baru.\n\nPerkiraan trafik dalam 1 menit terakhir: 30 + (80 × 0.75) = 30 + 60 = 90 request.\n\nKarena 90 masih di bawah 100, *request*-nya dilolosin.\n\nPendekatan ini akurat banget, masalah *boundary spike*-nya hilang, dan cuma butuh penyimpanan dua angka *counter* per klien. Bandingin sama *Sliding Window Log* yang harus nyimpen *timestamp* tiap *request*—jauh lebih hemat.\n\nJadi gimana, udah mulai kebayang bedanya? Belum selesai, masih ada bagian serunya nih. 😎\n\n## Kapan pilih yang mana?\n\n| Algoritma | Toleransi *Burst* | Efisiensi Memori | Kompleksitas | *Use Case* Paling Pas |\n| :--- | :--- | :--- | :--- | :--- |\n| **Token Bucket** | ⭐⭐⭐ Sangat Bagus | ⭐⭐⭐ Sangat Ringan | Menengah | Public API, SaaS Gateway, REST API umum |\n| **Leaky Bucket** | ❌ Nggak Ada (Smooth) | ⭐⭐ Sedang (*Queue*) | Menengah | Antrean ke *third-party payment gateway* |\n| **Fixed Window** | ❌ Rawan Lonjakan Batas | ⭐⭐⭐ Sangat Ringan | Sangat Mudah | Proteksi kasar internal / *cron jobs* |\n| **Sliding Window** | ⭐⭐ Cukup Baik | ⭐⭐⭐ Sangat Ringan | Menengah | API kuota per jam/hari dengan batas ketat |\n\n## Implementasi praktis: token bucket dengan Redis dan Lua\n\nDi sistem *microservices* terdistribusi, kita wajib pakai *shared store* kayak [Redis](https://redis.io/) biar kuota klien tetap sinkron lintas semua pod *instance API Gateway*.\n\nBiar nggak ada *race condition*—dua *request* baca token barengan terus dua-duanya lolos—seluruh logika cek dan ambil token harus jalan atomik lewat **Lua Script**:\n\n```lua\n-- Redis Lua Script untuk Token Bucket\nlocal key = KEYS[1]\nlocal limit = tonumber(ARGV[1])        -- Kapasitas maksimum ember (cth: 10)\nlocal current_time = tonumber(ARGV[2]) -- Timestamp saat ini dalam detik\nlocal refill_rate = tonumber(ARGV[3])  -- Token per detik (cth: 2)\n\n-- Ambil data bucket: [tokens, last_updated]\nlocal data = redis.call(\"HMGET\", key, \"tokens\", \"last_updated\")\nlocal tokens = tonumber(data[1])\nlocal last_updated = tonumber(data[2])\n\nif tokens == nil then\n    tokens = limit\n    last_updated = current_time\nelse\n    -- Hitung penambahan token sejak request terakhir\n    local delta = math.max(0, current_time - last_updated)\n    tokens = math.min(limit, tokens + (delta * refill_rate))\n    last_updated = current_time\nend\n\n-- Periksa apakah token mencukupi\nif tokens >= 1 then\n    tokens = tokens - 1\n    redis.call(\"HSET\", key, \"tokens\", tokens, \"last_updated\", last_updated)\n    redis.call(\"EXPIRE\", key, 3600) -- TTL 1 jam\n    return 1 -- Izinkan request\nelse\n    redis.call(\"HSET\", key, \"tokens\", tokens, \"last_updated\", last_updated)\n    return 0 -- Tolak request (429)\nend\n```\n\nIntegrasi sederhananya di aplikasi *backend* (FastAPI / Python):\n\n```python\nimport time\nimport redis\nfrom fastapi import FastAPI, HTTPException, Request\n\napp = FastAPI()\nr = redis.Redis(host=\"localhost\", port=6379, db=0)\n\n# Load script Lua ke Redis\nrate_limit_lua = r.register_script(open(\"token_bucket.lua\").read())\n\n@app.middleware(\"http\")\nasync def rate_limit_middleware(request: Request, call_next):\n    client_ip = request.client.host\n    key = f\"rate_limit:{client_ip}\"\n\n    # Kapasitas 10 token, refill 2 token per detik\n    allowed = rate_limit_lua(keys=[key], args=[10, int(time.time()), 2])\n\n    if not allowed:\n        raise HTTPException(status_code=429, detail=\"Too Many Requests. Pelan-pelan ya!\")\n\n    return await call_next(request)\n```\n\n## Jadi, algoritma mana yang cocok buat kalian?\n\nPada akhirnya, nggak ada algoritma yang menang di semua situasi. Token Bucket juara buat API publik karena ramah *burst*. Leaky Bucket menang kalau kalian butuh aliran yang benar-benar rata. Fixed Window cuma cocok buat proteksi kasar yang nggak butuh presisi tinggi. Sliding Window Counter jadi kompromi paling seimbang buat kuota ketat per jam atau per hari.\n\nYang paling penting, hitungan kuotanya disimpan di *shared store* dan diperiksa secara atomik. Soalnya *rate limiter* yang salah hitung gara-gara *race condition* itu kayak kasir yang ngitung duit sambil merem—kelihatan jalan, tapi saldonya udah kacau dari awal.\n\nSelamat memilih algoritma ria, dan semoga *bucket* kalian nggak pernah bocor pas jam sibuk! 👋",[62],{"authors_id":63},{"name":64,"slug":65,"role":66,"profile_picture":58},"Inva","inva","Writer",{"id":68,"title":8,"slug":9,"articles":69,"parent_series":58},"9",[70,84,87,99,110,123,135,146,158,169,180,190,200,211,220,231,242,252,260,271,282,293,301,310,319,332,341,352,363,373,382,391,401],{"id":71,"title":72,"slug":73,"excerpt":74,"thumbnail_image":75,"tags":77},"41","Log Aggregation Modern dengan Grafana Loki dan Vector: Hemat Storage Tanpa Indeks Raksasa","log-aggregation-modern-grafana-loki-vector-hemat-storage","Klaster Elasticsearch kalian rakus RAM dan boros storage, ya? Padahal nggak semua kata di log perlu diindeks, lho. Yuk, kita kenalan sama duet Grafana Loki dan Vector yang jauh lebih ramah kantong!",{"id":76},"03c9d1a0-c80a-401d-9f2b-2b3ee7abb115",[51,78,79,80,81,82,83,52],"observa","logging","grafana","loki","vector","devops",{"id":43,"title":44,"slug":45,"excerpt":46,"thumbnail_image":85,"tags":86},{"id":49},[51,52,53,54,55,56,57],{"id":88,"title":89,"slug":90,"excerpt":91,"thumbnail_image":92,"tags":94},"16","Menggunakan TypeScript dengan Node.js dan Express","menggunakan-typescript-dengan-node-js-dan-express","Pengen bikin server Express di Node.js tapi pakai fitur-fitur keren TypeScript? Yuk, intip cara setup lengkapnya dari nol di sini!",{"id":93},"798b60f1-9ef5-4f73-8d43-85825fd86acc",[95,96,97,98],"javascript","nodejs","express","typescript",{"id":100,"title":101,"slug":102,"excerpt":103,"thumbnail_image":104,"tags":106},"17","Pemrograman Asynchronous vs Multithreading, Apa Sih Bedanya?","pemrograman-asynchronous-dan-multithreading-apa-bedanya","Sering dengar istilah asynchronous dan multithreading pas ngoding tapi masih bingung bedanya? Yuk, kita bedah perbedaannya lewat analogi bikin mi instan yang gampang dipahami!",{"id":105},"9c05d0ae-e829-4f1b-b9f7-a84f3e0439d3",[107,108,109],"pemrograman","concurrency","tips-coding",{"id":111,"title":112,"slug":113,"excerpt":114,"thumbnail_image":115,"tags":117},"33","Berkenalan dengan RabbitMQ: Si Kurir Pesan Serba Bisa","berkenalan-dengan-rabbitmq-si-kurir-pesan-serba-bisa","Pernah nggak sih kalian klik Bayar Sekarang dan notifikasi suksesnya muncul seketika, padahal email konfirmasi baru nyampe beberapa detik kemudian? Rahasianya ada di satu kurir tak terlihat. Yuk, kita kenalan sama RabbitMQ!",{"id":116},"f6620f79-b3b2-4f9c-8104-283a335e06c4",[118,119,51,120,121,122],"rabbitmq","message-broker","microservices","async","queue",{"id":124,"title":125,"slug":126,"excerpt":127,"thumbnail_image":128,"tags":130},"34","Kenapa Backend Butuh Database Connection Pooling?","kenapa-backend-butuh-database-connection-pooling","Pernah kena error Too many connections padahal CPU server masih santai di angka 30%? Masalahnya bukan di query kalian, tapi di cara aplikasi ngobrol sama database. Yuk, kenalan sama Connection Pooling!",{"id":129},"6f7a2d8c-2a37-4901-a522-50df8d9c0a20",[51,131,132,133,52,134],"database","postgresql","connection-pooling","performance",{"id":136,"title":137,"slug":138,"excerpt":139,"thumbnail_image":140,"tags":142},"27","Background Worker & Task Queue: Jangan Jalankan Proses Berat di Request HTTP!","background-worker-task-queue-jangan-jalankan-proses-berat-di-http","Pernah nggak sih kalian klik Daftar Akun lalu halaman loading muter sampai 15 detik tanpa kepastian? Penyebabnya klasik banget. Yuk, kita bahas kenapa proses berat nggak boleh dijalanin di request HTTP!",{"id":141},"e119c6ad-fc37-4b92-abdc-3eb20f8e34b4",[51,143,144,145,118,54,121],"task-queue","celery","fastapi",{"id":147,"title":148,"slug":149,"excerpt":150,"thumbnail_image":151,"tags":153},"28","Mengatasi N+1 Query Problem: Musuh Tersembunyi Performa Backend","mengatasi-n-plus-1-query-problem-musuh-tersembunyi-backend","Pernah bikin endpoint yang cuma nampilin 50 postingan beserta nama penulisnya, tapi response-nya lambat banget? Bisa jadi kalian kena N+1 Query. Yuk, kita bedah penyebabnya dan cara memberantasnya!",{"id":152},"34b94e14-3e9a-4778-9f1f-6df5898915fb",[131,154,155,156,157,134,51],"sql","orm","sqlalchemy","django",{"id":159,"title":160,"slug":161,"excerpt":162,"thumbnail_image":163,"tags":165},"30","Berkenalan dengan Bahasa Go: Kenapa Banyak Dipakai untuk Backend Modern?","berkenalan-dengan-bahasa-go-untuk-backend-modern","Kok bisa sih Go yang masih muda ini jadi bahasa andalan Docker sampai Kubernetes? Yuk, kita kenalan sama Goroutine, Channel, dan alasan dia hemat memori luar biasa!",{"id":164},"a368631c-f331-4c50-8322-1d0fc44927aa",[166,167,51,108,168,52],"golang","go","goroutine",{"id":170,"title":171,"slug":172,"excerpt":173,"thumbnail_image":174,"tags":176},"13","Hal yang Perlu Diperhatikan Sebelum Pindah ke Linux","hal-yang-perlu-diperhatikan-sebelum-pindah-ke-linux","Tertarik migrasi dari Windows ke Linux tapi masih ragu? Yuk, pelajari kelebihan, kekurangan, dan persiapan penting sebelum kamu benar-benar pindah!",{"id":175},"ae27713e-5025-4d13-88b6-efb4860cc5b3",[177,178,179],"linux","windows","foss",{"id":181,"title":182,"slug":183,"excerpt":184,"thumbnail_image":185,"tags":187},"11","Apa Kabar JavaScript? Rilis State of JS 2022","apa-kabar-javascript-rilis-state-of-js-2022","Survei State of JS 2022 resmi dirilis dengan lonjakan responden yang luar biasa! Framework dan build tool apa saja yang berhasil mencuri perhatian tahun ini?",{"id":186},"18df5262-d813-43d1-b1f7-5595a7f53602",[95,188,189],"web","news",{"id":191,"title":192,"slug":193,"excerpt":194,"thumbnail_image":195,"tags":197},"12","Buat PowerShell Lebih Berwarna dengan Starship","buat-powershell-lebih-berwarna-dengan-starship","Bosen sama tampilan PowerShell yang hitam putih dan kaku? Yuk, sulap terminalmu jadi lebih interaktif, estetik, dan berwarna pakai Starship!",{"id":196},"c5a8262b-43db-4479-8f6c-5a6a0d61653b",[198,178,199],"shell","powershell",{"id":201,"title":202,"slug":203,"excerpt":204,"thumbnail_image":205,"tags":207},"14","Instalasi Java dengan Visual Studio Code","instalasi-java-dengan-visual-studio-code","Mau belajar ngoding Java tapi spek laptop pas-pasan buat buka IDE berat? Yuk, cari tahu cara setup VS Code biar ngoding Java tetap lancar dan enteng!",{"id":206},"99b01cc2-11ac-4449-9927-a1a7116347df",[208,209,210],"howto","java","editor",{"id":212,"title":213,"slug":214,"excerpt":215,"thumbnail_image":216,"tags":218},"15","Kenapa Ada Banyak Sekali Distro Linux?","kenapa-ada-banyak-sekali-distro-linux","Pernah bingung pas mau install Linux tapi malah disuguhi puluhan pilihan distro? Tenang, kalian nggak sendirian. Yuk, cari tahu kenapa distro Linux bisa sebanyak itu dan apa bedanya!",{"id":217},"b5352a4b-f274-4489-aca9-36814947d514",[177,219],"distro",{"id":221,"title":222,"slug":223,"excerpt":224,"thumbnail_image":225,"tags":227},"18","React, Angular, atau Vue: Pilih Framework yang Mana?","react-angular-vue-pilih-framework-yang-mana","Bingung menentukan pilihan di antara React, Angular, dan Vue untuk proyek web kalian berikutnya? Tenang, yuk kita bandingkan kelebihan masing-masing di sini!",{"id":226},"ff111270-80c2-4ccb-83d7-d5ab8c021201",[228,229,230],"web-development","frontend","framework",{"id":232,"title":233,"slug":234,"excerpt":235,"thumbnail_image":236,"tags":238},"19","Static Typing dan Dynamic Typing: Apa Bedanya?","static-typing-dan-dynamic-typing-apa-bedanya","Pernah bingung kenapa di Java tipe data harus dideklarasikan, sementara di Python tinggal pakai aja? Yuk, cari tahu perbedaan static typing dan dynamic typing di sini!",{"id":237},"98e85e76-9e7b-42d1-b989-7c75e7f49abc",[239,240,241],"programming","types","comparison",{"id":243,"title":244,"slug":245,"excerpt":246,"thumbnail_image":247,"tags":249},"20","Tailwind v3.0 Rilis: Ada JIT, Arbitrary Values, dan Banyak Lagi!","tailwind-v3-0-rilis-jit-arbitrary-dan-banyak-lagi","Tailwind CSS v3.0 resmi dirilis dengan segudang peningkatan performa yang bikin proses coding makin asyik. Apa aja sih fitur andalan yang wajib dicoba?",{"id":248},"d5e879fa-5cd0-4015-a155-b0cc69ff12fb",[250,251,228],"tailwind-css","css-framework",{"id":253,"title":254,"slug":255,"excerpt":256,"thumbnail_image":257,"tags":259},"21","TypeScript: JavaScript dengan Gaya dan Tipe Data Statis","typescript-javascript-dengan-gaya","Pernah pusing gara-gara runtime error di JavaScript? Yuk, kenalan dengan TypeScript yang bawa fitur static typing biar kode kalian makin rapi dan aman!",{"id":258},"f615bb1f-8e36-4d6c-9c2c-f1629407173c",[98,95,228],{"id":261,"title":262,"slug":263,"excerpt":264,"thumbnail_image":265,"tags":267},"22","Yang Baru di Nuxt 3: Nitro Engine, Vite, dan Fitur Keren Lainnya","yang-baru-di-nuxt-3-nitro-vite-dan-banyak-lagi","Nuxt 3 hadir membawa perubahan arsitektur besar-besaran untuk menyelaraskan dengan Vue 3. Penasaran apa saja peningkatan performa dan fitur barunya?",{"id":266},"8444bdbc-f60c-4d8b-8593-1b49400e8659",[268,269,270],"nuxt-3","vue-3","frontend-framework",{"id":272,"title":273,"slug":274,"excerpt":275,"thumbnail_image":276,"tags":278},"23","Kenapa Pakai FastAPI dan Kenapa Dia Kenceng Banget?","kenapa-pakai-fastapi-dan-kenapa-dia-kenceng-banget","Sering dengar FastAPI tapi masih bingung kenapa dia bisa ngebut kayak NodeJS dan Go? Yuk, kita bedah bareng rahasia ASGI, sihir Pydantic, sampai dokumentasi otomatisnya!",{"id":277},"84847c9e-70ab-4dcc-bed7-7be50165f7f4",[145,279,51,280,281],"python","asgi","pydantic",{"id":283,"title":284,"slug":285,"excerpt":286,"thumbnail_image":287,"tags":289},"26","REST API vs gRPC: Kapan Sebaiknya Mulai Pindah dari JSON?","rest-api-vs-grpc-kapan-sebaiknya-pindah-dari-json","Pernah nggak sih sistem mikroservis kalian mulai kerasa berat padahal cuma tuker-tukeran data internal? Belum tentu masalah servernya. Yuk, kita bahas kapan REST + JSON masih juara dan kapan gRPC jadi jawabannya!",{"id":288},"487500a0-9e09-469d-8736-1583c5ea83aa",[51,290,291,120,292,52],"grpc","rest-api","protobuf",{"id":294,"title":295,"slug":296,"excerpt":297,"thumbnail_image":298,"tags":300},"38","Optimistic vs Pessimistic Locking: Mengatasi Rebutan Data di Database","optimistic-vs-pessimistic-locking-mengatasi-rebutan-data-di-database","Seribu orang ngeklik \"Beli Sekarang\" di detik yang sama buat rebutan 1 tiket terakhir. Kira-kira apa yang bakal terjadi di server kalian? Yuk, kita banding *pessimistic* sama *optimistic locking*!",{"id":299},"21ab769d-6f82-4a22-be11-17ac16e9fb6b",[51,131,108,132,52,56,154],{"id":302,"title":303,"slug":304,"excerpt":305,"thumbnail_image":306,"tags":308},"24","Kenapa Butuh Rate Limiting?","kenapa-butuh-rate-limiting","Pernah nggak sih endpoint login kalian diserbu bot yang nyoba nembak ribuan password cuma dalam semenit? Yuk, kita bahas kenapa Rate Limiting itu wajib, plus cara masangnya di FastAPI!",{"id":307},"1401f9e5-6359-404a-8d80-273ddee9a372",[51,53,52,309,54,145],"api",{"id":311,"title":312,"slug":313,"excerpt":314,"thumbnail_image":315,"tags":317},"25","Kenapa Event Queue Bikin Backend Kalian Nggak Gampang Tumbang?","kenapa-event-queue-bikin-backend-kamu-nggak-gampang-tumbang","Pernah klik tombol Beli Sekarang dan notifikasi suksesnya muncul dalam sekejap, padahal di belakang layar ada invoice, email, sampai push notification yang harus dibuat? Rahasianya Event Queue. Yuk, kita bedah!",{"id":316},"d9e76018-fbcc-4c69-ad42-8f75ff229ecd",[51,52,318,119,56],"event-queue",{"id":320,"title":321,"slug":322,"excerpt":323,"thumbnail_image":324,"tags":326},"46","Mengapa C Mulai Ditinggalkan: Duel Filosofi Keamanan Rust vs Kesederhanaan Zig","mengapa-c-mulai-ditinggalkan-duel-filosofi-rust-vs-zig","Pernah denger kalau sekitar 70% celah keamanan kritis software modern itu asalnya dari bahasa C? Iya, bahasa yang udah jadi fondasi internet ini 😅. Yuk, kita bedah duel filosofi Rust vs Zig dan cari tahu mana yang pas buat proyek kalian!",{"id":325},"1942a5c9-c1f6-476d-9913-768f78d9d8eb",[327,328,329,330,331],"system-programming","rust","zig","c","memory-safety",{"id":333,"title":334,"slug":335,"excerpt":336,"thumbnail_image":337,"tags":339},"29","Memahami Database Indexing: Rahasia Query Super Cepat","memahami-database-indexing-rahasia-query-super-cepat","Query kalian mulai lelet padahal data baru nembus jutaan baris? Sebelum buru-buru *upgrade* server, yuk cek dulu *index*-nya. Yuk, kita bedah gimana *database indexing* sebenernya kerja!",{"id":338},"b123036f-1779-478c-a545-cf5f189b5fe9",[131,154,132,340,134,51],"indexing",{"id":342,"title":343,"slug":344,"excerpt":345,"thumbnail_image":346,"tags":348},"31","Mengenal Redis: Lebih dari Sekadar In-Memory Cache","mengenal-redis-lebih-dari-sekadar-in-memory-cache","Masih nganggep Redis cuma buat caching? Padahal dia bisa jadi leaderboard, message broker, sampai rate limiter lho. Yuk, kita bedah semua kemampuannya!",{"id":347},"9595b7d7-60e0-48f5-a715-f3cf60f34e1d",[54,51,349,350,351,279],"cache","nosql","in-memory",{"id":353,"title":354,"slug":355,"excerpt":356,"thumbnail_image":357,"tags":359},"32","Docker Compose untuk Developer Santai: Satu Perintah Buat Jalankan Semua Service","docker-compose-untuk-developer-santai","Pernah nggak kalian ngulang-ngulang perintah `docker run` cuma buat nyalain web, database, dan cache satu-satu? Capek, kan? Yuk, kita rapikan semuanya jadi satu file dan satu perintah!",{"id":358},"f4c08f1b-e0d0-410c-b091-c3618944fb3e",[360,361,83,362,51],"docker","docker-compose","containers",{"id":364,"title":365,"slug":366,"excerpt":367,"thumbnail_image":368,"tags":370},"35","Mengenal Idempotency Key: Rahasia Anti-Double Charge di API Pembayaran","mengenal-idempotency-key-rahasia-anti-double-charge-di-api-pembayaran","Pernah internet kalian ngadat pas lagi nekan tombol Bayar, terus panik ngekliknya tiga kali? Kalau endpoint-nya nggak idempotent, saldo bisa kepotong berkali-kali. Yuk, kenalan sama Idempotency Key!",{"id":369},"724cb83a-3ce5-4ee2-b0d7-533e3a0fc7df",[51,309,371,52,372,145,54],"idempotency","payment",{"id":374,"title":375,"slug":376,"excerpt":377,"thumbnail_image":378,"tags":380},"36","Mencegah Efek Domino di Mikroservis dengan Circuit Breaker Pattern","mencegah-efek-domino-di-mikroservis-dengan-circuit-breaker-pattern","Pernah nggak sih, satu fitur minor di aplikasi kalian mendadak lelet, eh malah bikin halaman checkout dan login ikut down total? Itu namanya *cascading failure*. Yuk, kita bahas gimana *circuit breaker pattern* (pemutus arus) nyelametin arsitektur kalian!",{"id":379},"c2e8e37f-8f9a-423f-a09d-74613b3c2924",[51,120,381,52,56],"circuit-breaker",{"id":383,"title":384,"slug":385,"excerpt":386,"thumbnail_image":387,"tags":389},"37","Kenapa Backend Butuh Graceful Shutdown? Jangan Asal Cabut Kabel Server!","kenapa-backend-butuh-graceful-shutdown-jangan-asal-cabut-kabel-server","Pernah nggak sih, abis kalian nge-*deploy* backend, beberapa detik kemudian monitoring malah banjir *alert* **502 Bad Gateway** dari pengguna? Saldo udah kepotong tapi status order belum keburu ter-*update* di *database*. Yuk, kita bahas kenapa *graceful shutdown* itu wajib banget!",{"id":388},"251a54a7-1d24-47fd-8cbc-a46fd42cbb6f",[51,83,145,360,390,52,56],"kubernetes",{"id":392,"title":393,"slug":394,"excerpt":395,"thumbnail_image":396,"tags":398},"39","Database Sharding vs Partitioning: Kapan Data Perlu Dipecah?","database-sharding-vs-partitioning-kapan-data-perlu-dipecah","Tabel transaksi kalian udah nembus miliaran baris dan tiap *query* laporan bikin *CPU* melonjak 100%? Sebelum asal pecah data, yuk kita pahami dulu beda *sharding* sama *partitioning*!",{"id":397},"08ef5fe5-d935-47b3-bbbc-d976948b2a9e",[51,131,132,52,56,399,400],"sharding","scaling",{"id":402,"title":403,"slug":404,"excerpt":405,"thumbnail_image":406,"tags":408},"40","Menyelami Distributed Tracing: Melacak Jejak Request di Balik Keruwetan Microservices","menyelami-distributed-tracing-melacak-jejak-request-microservices","Satu klik Checkout di backend kalian bisa memicu puluhan network call antar-service, tapi pas ada yang lambat, service mana sebenernya yang salah? Log aja nggak cukup buat njawab itu. Yuk, kita bedah Distributed Tracing!",{"id":407},"1ace92fa-b89c-4cef-ab75-4e4cd67a089c",[51,409,410,411,120,56,412],"distributed-systems","observability","open-telemetry","tracing","\u003Cp>Di artikel sebelumnya kita udah ngobrol soal kenapa \u003Cem>rate limiting\u003C/em> itu penting banget. Nah, sekarang kita masuk ke bagian yang lebih seru: algoritmanya sendiri. 🤔\u003C/p>\n\u003Cp>Nyatanya, bikin \u003Cem>rate limiter\u003C/em> itu bukan sekadar ngecek \u003Ccode>request_count > 100\u003C/code>. Di balik setiap \u003Cem>API Gateway\u003C/em> modern—kayak \u003Ca href=\"https://konghq.com/\">Kong\u003C/a>, \u003Ca href=\"https://nginx.org/\">NGINX\u003C/a>, atau Cloudflare—ada perhitungan matematis yang nentuin gimana trafik kalian diperlakukan.\u003C/p>\n\u003Cp>Ada empat algoritma yang paling sering dipakai di industri. Karakternya beda-beda, dan salah milih bisa bikin kuota kalian bocor atau pengguna asli malah kena blokir. Yuk, kita bedah satu-satu!\u003C/p>\n\u003Ch2 id=\"token-bucket-si-favorit-yang-fleksibel\">Token bucket, si favorit yang fleksibel\u003C/h2>\n\u003Cp>\u003Cstrong>Token Bucket\u003C/strong> jadi standar \u003Cem>de facto\u003C/em> di sebagian besar \u003Cem>API Gateway\u003C/em> modern, termasuk AWS API Gateway dan Google Cloud Armor.\u003C/p>\n\u003Cp>Cara kerjanya gampang. Bayangin satu ember dengan kapasitas token terbatas, misalnya 10 token.\u003C/p>\n\u003Cul>\n\u003Cli>Tiap detik, sistem nambah token baru ke ember dengan laju konstan (misal 2 token/detik). Kalau embernya udah penuh, token barunya tumpah dan dibuang percuma.\u003C/li>\n\u003Cli>Setiap \u003Cem>request\u003C/em> HTTP dari klien harus ngambil 1 token biar boleh lewat.\u003C/li>\n\u003Cli>Token masih ada? Silakan diproses. Token habis? \u003Cem>Request\u003C/em>-nya langsung ditolak dengan \u003Ccode>429 Too Many Requests\u003C/code>.\u003C/li>\n\u003C/ul>\n\u003Cp>Enaknya, algoritma ini ramah banget sama \u003Cem>burst\u003C/em>. Kalau klien diem beberapa detik, embernya ngisi penuh. Dia bisa nembak 10 \u003Cem>request\u003C/em> sekaligus dalam 1 milidetik tanpa ditolak, selama rata-rata jangka panjangnya tetap sesuai kuota.\u003C/p>\n\u003Cp>\u003Cem>Trade-off\u003C/em>-nya: kalian mesti \u003Cem>tuning\u003C/em> dua parameter sekaligus. \u003Cem>Bucket size\u003C/em> nentuin seberapa besar \u003Cem>burst\u003C/em> yang ditoleransi, \u003Cem>refill rate\u003C/em> nentuin seberapa cepat kuotanya balik. Salah angka dikit, limiternya bisa kerasa terlalu longgar atau malah terlalu galak.\u003C/p>\n\u003Ch2 id=\"leaky-bucket-yang-doyan-ngerapiin-antrean\">Leaky bucket, yang doyan ngerapiin antrean\u003C/h2>\n\u003Cp>Kalau Token Bucket ngatur jumlah token, \u003Cstrong>Leaky Bucket\u003C/strong> ngatur laju keluaran \u003Cem>request\u003C/em> (\u003Cem>throughput smoothing\u003C/em>).\u003C/p>\n\u003Cp>Ibaratnya ember berlubang kecil di dasarnya.\u003C/p>\n\u003Cul>\n\u003Cli>\u003Cem>Request\u003C/em> dari klien masuk kayak air dituang dari atas, ditampung di antrean FIFO.\u003C/li>\n\u003Cli>Air netes keluar dari lubang dasar dengan laju stabil, misal tepat 5 \u003Cem>request\u003C/em> per detik, seberapa pun derasnya tuangan di atas.\u003C/li>\n\u003Cli>Kalau tuangannya kebanyakan sampai embernya meluap, air yang luber langsung dibuang—\u003Cem>request\u003C/em>-nya ditolak dengan \u003Ccode>429\u003C/code>.\u003C/li>\n\u003C/ul>\n\u003Cp>Kelebihannya: alirannya jadi sangat stabil (\u003Cem>constant rate\u003C/em>). Cocok banget dipasang di depan \u003Cem>service\u003C/em> yang sensitif ke beban mendadak, misal sistem pembayaran perbankan warisan (\u003Cem>legacy system\u003C/em>) yang nggak sanggup nerima lonjakan.\u003C/p>\n\u003Cp>Kekurangannya keras: dia nggak toleran sama \u003Cem>burst\u003C/em> sama sekali. \u003Cem>Request\u003C/em> yang sebenarnya sah malah nambah latensi karena harus ngantre di \u003Cem>buffer\u003C/em>, padahal server kalian masih sanggup kok ngelayanin.\u003C/p>\n\u003Ch2 id=\"fixed-window-counter-simpel-tapi-ada-jebakannya\">Fixed window counter, simpel tapi ada jebakannya\u003C/h2>\n\u003Cp>Ini algoritma paling gampang dan paling sering ditulis \u003Cem>developer\u003C/em> pemula.\u003C/p>\n\u003Cp>Waktu dibagi jadi jendela-jendela tetap (\u003Cem>fixed time window\u003C/em>), misal per menit: 00:00-00:01, 00:01-00:02, dan seterusnya.\u003C/p>\n\u003Cul>\n\u003Cli>Tiap \u003Cem>request\u003C/em> di interval itu nambah \u003Cem>counter\u003C/em> sebesar 1.\u003C/li>\n\u003Cli>Kalau \u003Cem>counter\u003C/em>-nya lewat ambang batas (misal 100 req/menit), \u003Cem>request\u003C/em> berikutnya ditolak sampai jendelanya berganti dan \u003Cem>counter\u003C/em>-nya di-reset ke 0.\u003C/li>\n\u003C/ul>\n\u003Cp>Masalahnya ada di jahitan antar-jendela. Anggap kuotanya 100 req/menit:\u003C/p>\n\u003Cul>\n\u003Cli>Klien nembak 100 \u003Cem>request\u003C/em> di detik 00:59. Lolos semua.\u003C/li>\n\u003Cli>Klien nembak 100 \u003Cem>request\u003C/em> lagi di detik 01:01. Lolos juga, soalnya \u003Cem>counter\u003C/em>-nya baru di-reset.\u003C/li>\n\u003C/ul>\n\u003Cp>Hasilnya, server nerima \u003Cstrong>200 \u003Cem>request\u003C/em> cuma dalam rentang 2 detik\u003C/strong>! Kapasitas yang kalian rencanain buat satu menit penuh, dilahap habis dalam sekejap. Inilah yang namanya \u003Cem>boundary spike\u003C/em>, dan pas lagi \u003Cem>flash sale\u003C/em>, dia bisa bikin \u003Cem>backend\u003C/em> kalian tumbang. 😅\u003C/p>\n\u003Ch2 id=\"sliding-window-counter-akurat-tanpa-boros-memori\">Sliding window counter, akurat tanpa boros memori\u003C/h2>\n\u003Cp>Buat nutup celah \u003Cem>boundary spike\u003C/em> tanpa harus nyimpen \u003Cem>log\u003C/em> tiap \u003Cem>request\u003C/em>—itu namanya \u003Cem>Sliding Window Log\u003C/em>, dan boros memori—industri pindah ke pendekatan \u003Cstrong>Sliding Window Counter\u003C/strong>.\u003C/p>\n\u003Cp>Idenya: hitung estimasi \u003Cem>request\u003C/em> di jendela geser saat ini, dengan nimbrung proporsi dari jendela sebelumnya.\u003C/p>\n\u003Cp>$$\\text{Estimated Requests} = \\text{Count}\u003Cem>{\\text{current}} + \\left( \\text{Count}\u003C/em>{\\text{previous}} \\times (1 - \\text{Overlap Ratio}) \\right)$$\u003C/p>\n\u003Cp>Contoh, kuota 100 req/menit:\u003C/p>\n\u003Cul>\n\u003Cli>Jendela sebelumnya (00:00-01:00) nyatet 80 \u003Cem>request\u003C/em>.\u003C/li>\n\u003Cli>Sekarang jam 01:15, artinya kita baru 25% masuk ke jendela saat ini. Sisanya 75% masih nempel ke jendela sebelumnya.\u003C/li>\n\u003Cli>Di jendela saat ini udah ada 30 \u003Cem>request\u003C/em> baru.\u003C/li>\n\u003C/ul>\n\u003Cp>Perkiraan trafik dalam 1 menit terakhir: 30 + (80 × 0.75) = 30 + 60 = 90 request.\u003C/p>\n\u003Cp>Karena 90 masih di bawah 100, \u003Cem>request\u003C/em>-nya dilolosin.\u003C/p>\n\u003Cp>Pendekatan ini akurat banget, masalah \u003Cem>boundary spike\u003C/em>-nya hilang, dan cuma butuh penyimpanan dua angka \u003Cem>counter\u003C/em> per klien. Bandingin sama \u003Cem>Sliding Window Log\u003C/em> yang harus nyimpen \u003Cem>timestamp\u003C/em> tiap \u003Cem>request\u003C/em>—jauh lebih hemat.\u003C/p>\n\u003Cp>Jadi gimana, udah mulai kebayang bedanya? Belum selesai, masih ada bagian serunya nih. 😎\u003C/p>\n\u003Ch2 id=\"kapan-pilih-yang-mana\">Kapan pilih yang mana?\u003C/h2>\n\u003Ctable>\n\u003Cthead>\n\u003Ctr>\n\u003Cth align=\"left\">Algoritma\u003C/th>\n\u003Cth align=\"left\">Toleransi \u003Cem>Burst\u003C/em>\u003C/th>\n\u003Cth align=\"left\">Efisiensi Memori\u003C/th>\n\u003Cth align=\"left\">Kompleksitas\u003C/th>\n\u003Cth align=\"left\">\u003Cem>Use Case\u003C/em> Paling Pas\u003C/th>\n\u003C/tr>\n\u003C/thead>\n\u003Ctbody>\u003Ctr>\n\u003Ctd align=\"left\">\u003Cstrong>Token Bucket\u003C/strong>\u003C/td>\n\u003Ctd align=\"left\">⭐⭐⭐ Sangat Bagus\u003C/td>\n\u003Ctd align=\"left\">⭐⭐⭐ Sangat Ringan\u003C/td>\n\u003Ctd align=\"left\">Menengah\u003C/td>\n\u003Ctd align=\"left\">Public API, SaaS Gateway, REST API umum\u003C/td>\n\u003C/tr>\n\u003Ctr>\n\u003Ctd align=\"left\">\u003Cstrong>Leaky Bucket\u003C/strong>\u003C/td>\n\u003Ctd align=\"left\">❌ Nggak Ada (Smooth)\u003C/td>\n\u003Ctd align=\"left\">⭐⭐ Sedang (\u003Cem>Queue\u003C/em>)\u003C/td>\n\u003Ctd align=\"left\">Menengah\u003C/td>\n\u003Ctd align=\"left\">Antrean ke \u003Cem>third-party payment gateway\u003C/em>\u003C/td>\n\u003C/tr>\n\u003Ctr>\n\u003Ctd align=\"left\">\u003Cstrong>Fixed Window\u003C/strong>\u003C/td>\n\u003Ctd align=\"left\">❌ Rawan Lonjakan Batas\u003C/td>\n\u003Ctd align=\"left\">⭐⭐⭐ Sangat Ringan\u003C/td>\n\u003Ctd align=\"left\">Sangat Mudah\u003C/td>\n\u003Ctd align=\"left\">Proteksi kasar internal / \u003Cem>cron jobs\u003C/em>\u003C/td>\n\u003C/tr>\n\u003Ctr>\n\u003Ctd align=\"left\">\u003Cstrong>Sliding Window\u003C/strong>\u003C/td>\n\u003Ctd align=\"left\">⭐⭐ Cukup Baik\u003C/td>\n\u003Ctd align=\"left\">⭐⭐⭐ Sangat Ringan\u003C/td>\n\u003Ctd align=\"left\">Menengah\u003C/td>\n\u003Ctd align=\"left\">API kuota per jam/hari dengan batas ketat\u003C/td>\n\u003C/tr>\n\u003C/tbody>\u003C/table>\n\u003Ch2 id=\"implementasi-praktis-token-bucket-dengan-redis-dan-lua\">Implementasi praktis: token bucket dengan Redis dan Lua\u003C/h2>\n\u003Cp>Di sistem \u003Cem>microservices\u003C/em> terdistribusi, kita wajib pakai \u003Cem>shared store\u003C/em> kayak \u003Ca href=\"https://redis.io/\">Redis\u003C/a> biar kuota klien tetap sinkron lintas semua pod \u003Cem>instance API Gateway\u003C/em>.\u003C/p>\n\u003Cp>Biar nggak ada \u003Cem>race condition\u003C/em>—dua \u003Cem>request\u003C/em> baca token barengan terus dua-duanya lolos—seluruh logika cek dan ambil token harus jalan atomik lewat \u003Cstrong>Lua Script\u003C/strong>:\u003C/p>\n\n\u003Cdiv class=\"code-block-wrapper my-6 not-prose rounded-xl border-2 border-black dark:border-gray-600 shadow-neo overflow-hidden bg-[#24292e] text-[#e1e4e8] transition-all\">\n  \u003Cdiv class=\"code-block-header flex items-center justify-between px-4 py-2.5 bg-[#1f2428] border-b-2 border-black dark:border-gray-600 font-mono text-xs font-bold text-gray-300 select-none\">\n    \u003Cdiv class=\"flex items-center gap-2\">\n      \u003Cspan class=\"w-3 h-3 rounded-full bg-[#ff5f56] border border-black/40 inline-block\">\u003C/span>\n      \u003Cspan class=\"w-3 h-3 rounded-full bg-[#ffbd2e] border border-black/40 inline-block\">\u003C/span>\n      \u003Cspan class=\"w-3 h-3 rounded-full bg-[#27c93f] border border-black/40 inline-block\">\u003C/span>\n      \u003Cspan class=\"ml-2 font-mono font-bold text-xs uppercase tracking-wider text-gray-300\">LUA\u003C/span>\n    \u003C/div>\n    \u003Cbutton type=\"button\" class=\"copy-code-btn flex items-center gap-1.5 px-3 py-1 text-xs font-bold bg-[#2f363d] text-gray-200 border-2 border-black dark:border-gray-500 rounded-lg shadow-neo-sm hover:translate-x-0.5 hover:translate-y-0.5 hover:shadow-none hover:bg-gray-700 transition-all cursor-pointer\" data-code=\"--%20Redis%20Lua%20Script%20untuk%20Token%20Bucket%0Alocal%20key%20%3D%20KEYS%5B1%5D%0Alocal%20limit%20%3D%20tonumber(ARGV%5B1%5D)%20%20%20%20%20%20%20%20--%20Kapasitas%20maksimum%20ember%20(cth%3A%2010)%0Alocal%20current_time%20%3D%20tonumber(ARGV%5B2%5D)%20--%20Timestamp%20saat%20ini%20dalam%20detik%0Alocal%20refill_rate%20%3D%20tonumber(ARGV%5B3%5D)%20%20--%20Token%20per%20detik%20(cth%3A%202)%0A%0A--%20Ambil%20data%20bucket%3A%20%5Btokens%2C%20last_updated%5D%0Alocal%20data%20%3D%20redis.call(%22HMGET%22%2C%20key%2C%20%22tokens%22%2C%20%22last_updated%22)%0Alocal%20tokens%20%3D%20tonumber(data%5B1%5D)%0Alocal%20last_updated%20%3D%20tonumber(data%5B2%5D)%0A%0Aif%20tokens%20%3D%3D%20nil%20then%0A%20%20%20%20tokens%20%3D%20limit%0A%20%20%20%20last_updated%20%3D%20current_time%0Aelse%0A%20%20%20%20--%20Hitung%20penambahan%20token%20sejak%20request%20terakhir%0A%20%20%20%20local%20delta%20%3D%20math.max(0%2C%20current_time%20-%20last_updated)%0A%20%20%20%20tokens%20%3D%20math.min(limit%2C%20tokens%20%2B%20(delta%20*%20refill_rate))%0A%20%20%20%20last_updated%20%3D%20current_time%0Aend%0A%0A--%20Periksa%20apakah%20token%20mencukupi%0Aif%20tokens%20%3E%3D%201%20then%0A%20%20%20%20tokens%20%3D%20tokens%20-%201%0A%20%20%20%20redis.call(%22HSET%22%2C%20key%2C%20%22tokens%22%2C%20tokens%2C%20%22last_updated%22%2C%20last_updated)%0A%20%20%20%20redis.call(%22EXPIRE%22%2C%20key%2C%203600)%20--%20TTL%201%20jam%0A%20%20%20%20return%201%20--%20Izinkan%20request%0Aelse%0A%20%20%20%20redis.call(%22HSET%22%2C%20key%2C%20%22tokens%22%2C%20tokens%2C%20%22last_updated%22%2C%20last_updated)%0A%20%20%20%20return%200%20--%20Tolak%20request%20(429)%0Aend\" title=\"Salin kode\">\n      \u003Csvg class=\"w-3.5 h-3.5 copy-icon-svg\" fill=\"none\" stroke=\"currentColor\" viewBox=\"0 0 24 24\">\n        \u003Cpath stroke-linecap=\"round\" stroke-linejoin=\"round\" stroke-width=\"2\" d=\"M8 16H6a2 2 0 01-2-2V6a2 2 0 012-2h8a2 2 0 012 2v2m-6 12h8a2 2 0 002-2v-8a2 2 0 00-2-2h-8a2 2 0 00-2 2v8a2 2 0 002 2z\">\u003C/path>\n      \u003C/svg>\n      \u003Cspan class=\"copy-text\">Copy\u003C/span>\n    \u003C/button>\n  \u003C/div>\n  \u003Cdiv class=\"code-block-content p-4 overflow-x-auto text-sm font-mono leading-relaxed bg-[#24292e]\">\n    \u003Cpre class=\"shiki github-dark\" style=\"background-color:#24292e;color:#e1e4e8\">\u003Ccode>-- Redis Lua Script untuk Token Bucket\nlocal key = KEYS[1]\nlocal limit = tonumber(ARGV[1])        -- Kapasitas maksimum ember (cth: 10)\nlocal current_time = tonumber(ARGV[2]) -- Timestamp saat ini dalam detik\nlocal refill_rate = tonumber(ARGV[3])  -- Token per detik (cth: 2)\n\n-- Ambil data bucket: [tokens, last_updated]\nlocal data = redis.call(\"HMGET\", key, \"tokens\", \"last_updated\")\nlocal tokens = tonumber(data[1])\nlocal last_updated = tonumber(data[2])\n\nif tokens == nil then\n    tokens = limit\n    last_updated = current_time\nelse\n    -- Hitung penambahan token sejak request terakhir\n    local delta = math.max(0, current_time - last_updated)\n    tokens = math.min(limit, tokens + (delta * refill_rate))\n    last_updated = current_time\nend\n\n-- Periksa apakah token mencukupi\nif tokens >= 1 then\n    tokens = tokens - 1\n    redis.call(\"HSET\", key, \"tokens\", tokens, \"last_updated\", last_updated)\n    redis.call(\"EXPIRE\", key, 3600) -- TTL 1 jam\n    return 1 -- Izinkan request\nelse\n    redis.call(\"HSET\", key, \"tokens\", tokens, \"last_updated\", last_updated)\n    return 0 -- Tolak request (429)\nend\u003C/code>\u003C/pre>\n  \u003C/div>\n\u003C/div>\u003Cp>Integrasi sederhananya di aplikasi \u003Cem>backend\u003C/em> (FastAPI / Python):\u003C/p>\n\n\u003Cdiv class=\"code-block-wrapper my-6 not-prose rounded-xl border-2 border-black dark:border-gray-600 shadow-neo overflow-hidden bg-[#24292e] text-[#e1e4e8] transition-all\">\n  \u003Cdiv class=\"code-block-header flex items-center justify-between px-4 py-2.5 bg-[#1f2428] border-b-2 border-black dark:border-gray-600 font-mono text-xs font-bold text-gray-300 select-none\">\n    \u003Cdiv class=\"flex items-center gap-2\">\n      \u003Cspan class=\"w-3 h-3 rounded-full bg-[#ff5f56] border border-black/40 inline-block\">\u003C/span>\n      \u003Cspan class=\"w-3 h-3 rounded-full bg-[#ffbd2e] border border-black/40 inline-block\">\u003C/span>\n      \u003Cspan class=\"w-3 h-3 rounded-full bg-[#27c93f] border border-black/40 inline-block\">\u003C/span>\n      \u003Cspan class=\"ml-2 font-mono font-bold text-xs uppercase tracking-wider text-gray-300\">PYTHON\u003C/span>\n    \u003C/div>\n    \u003Cbutton type=\"button\" class=\"copy-code-btn flex items-center gap-1.5 px-3 py-1 text-xs font-bold bg-[#2f363d] text-gray-200 border-2 border-black dark:border-gray-500 rounded-lg shadow-neo-sm hover:translate-x-0.5 hover:translate-y-0.5 hover:shadow-none hover:bg-gray-700 transition-all cursor-pointer\" data-code=\"import%20time%0Aimport%20redis%0Afrom%20fastapi%20import%20FastAPI%2C%20HTTPException%2C%20Request%0A%0Aapp%20%3D%20FastAPI()%0Ar%20%3D%20redis.Redis(host%3D%22localhost%22%2C%20port%3D6379%2C%20db%3D0)%0A%0A%23%20Load%20script%20Lua%20ke%20Redis%0Arate_limit_lua%20%3D%20r.register_script(open(%22token_bucket.lua%22).read())%0A%0A%40app.middleware(%22http%22)%0Aasync%20def%20rate_limit_middleware(request%3A%20Request%2C%20call_next)%3A%0A%20%20%20%20client_ip%20%3D%20request.client.host%0A%20%20%20%20key%20%3D%20f%22rate_limit%3A%7Bclient_ip%7D%22%0A%0A%20%20%20%20%23%20Kapasitas%2010%20token%2C%20refill%202%20token%20per%20detik%0A%20%20%20%20allowed%20%3D%20rate_limit_lua(keys%3D%5Bkey%5D%2C%20args%3D%5B10%2C%20int(time.time())%2C%202%5D)%0A%0A%20%20%20%20if%20not%20allowed%3A%0A%20%20%20%20%20%20%20%20raise%20HTTPException(status_code%3D429%2C%20detail%3D%22Too%20Many%20Requests.%20Pelan-pelan%20ya!%22)%0A%0A%20%20%20%20return%20await%20call_next(request)\" title=\"Salin kode\">\n      \u003Csvg class=\"w-3.5 h-3.5 copy-icon-svg\" fill=\"none\" stroke=\"currentColor\" viewBox=\"0 0 24 24\">\n        \u003Cpath stroke-linecap=\"round\" stroke-linejoin=\"round\" stroke-width=\"2\" d=\"M8 16H6a2 2 0 01-2-2V6a2 2 0 012-2h8a2 2 0 012 2v2m-6 12h8a2 2 0 002-2v-8a2 2 0 00-2-2h-8a2 2 0 00-2 2v8a2 2 0 002 2z\">\u003C/path>\n      \u003C/svg>\n      \u003Cspan class=\"copy-text\">Copy\u003C/span>\n    \u003C/button>\n  \u003C/div>\n  \u003Cdiv class=\"code-block-content p-4 overflow-x-auto text-sm font-mono leading-relaxed bg-[#24292e]\">\n    \u003Cpre class=\"shiki github-dark\" style=\"background-color:#24292e;color:#e1e4e8\" tabindex=\"0\">\u003Ccode>\u003Cspan class=\"line\">\u003Cspan style=\"color:#F97583\">import\u003C/span>\u003Cspan style=\"color:#E1E4E8\"> time\u003C/span>\u003C/span>\n\u003Cspan class=\"line\">\u003Cspan style=\"color:#F97583\">import\u003C/span>\u003Cspan style=\"color:#E1E4E8\"> redis\u003C/span>\u003C/span>\n\u003Cspan class=\"line\">\u003Cspan style=\"color:#F97583\">from\u003C/span>\u003Cspan style=\"color:#E1E4E8\"> fastapi \u003C/span>\u003Cspan style=\"color:#F97583\">import\u003C/span>\u003Cspan style=\"color:#E1E4E8\"> FastAPI, HTTPException, Request\u003C/span>\u003C/span>\n\u003Cspan class=\"line\">\u003C/span>\n\u003Cspan class=\"line\">\u003Cspan style=\"color:#E1E4E8\">app \u003C/span>\u003Cspan style=\"color:#F97583\">=\u003C/span>\u003Cspan style=\"color:#E1E4E8\"> FastAPI()\u003C/span>\u003C/span>\n\u003Cspan class=\"line\">\u003Cspan style=\"color:#E1E4E8\">r \u003C/span>\u003Cspan style=\"color:#F97583\">=\u003C/span>\u003Cspan style=\"color:#E1E4E8\"> redis.Redis(\u003C/span>\u003Cspan style=\"color:#FFAB70\">host\u003C/span>\u003Cspan style=\"color:#F97583\">=\u003C/span>\u003Cspan style=\"color:#9ECBFF\">\"localhost\"\u003C/span>\u003Cspan style=\"color:#E1E4E8\">, \u003C/span>\u003Cspan style=\"color:#FFAB70\">port\u003C/span>\u003Cspan style=\"color:#F97583\">=\u003C/span>\u003Cspan style=\"color:#79B8FF\">6379\u003C/span>\u003Cspan style=\"color:#E1E4E8\">, \u003C/span>\u003Cspan style=\"color:#FFAB70\">db\u003C/span>\u003Cspan style=\"color:#F97583\">=\u003C/span>\u003Cspan style=\"color:#79B8FF\">0\u003C/span>\u003Cspan style=\"color:#E1E4E8\">)\u003C/span>\u003C/span>\n\u003Cspan class=\"line\">\u003C/span>\n\u003Cspan class=\"line\">\u003Cspan style=\"color:#6A737D\"># Load script Lua ke Redis\u003C/span>\u003C/span>\n\u003Cspan class=\"line\">\u003Cspan style=\"color:#E1E4E8\">rate_limit_lua \u003C/span>\u003Cspan style=\"color:#F97583\">=\u003C/span>\u003Cspan style=\"color:#E1E4E8\"> r.register_script(\u003C/span>\u003Cspan style=\"color:#79B8FF\">open\u003C/span>\u003Cspan style=\"color:#E1E4E8\">(\u003C/span>\u003Cspan style=\"color:#9ECBFF\">\"token_bucket.lua\"\u003C/span>\u003Cspan style=\"color:#E1E4E8\">).read())\u003C/span>\u003C/span>\n\u003Cspan class=\"line\">\u003C/span>\n\u003Cspan class=\"line\">\u003Cspan style=\"color:#B392F0\">@app.middleware\u003C/span>\u003Cspan style=\"color:#E1E4E8\">(\u003C/span>\u003Cspan style=\"color:#9ECBFF\">\"http\"\u003C/span>\u003Cspan style=\"color:#E1E4E8\">)\u003C/span>\u003C/span>\n\u003Cspan class=\"line\">\u003Cspan style=\"color:#F97583\">async\u003C/span>\u003Cspan style=\"color:#F97583\"> def\u003C/span>\u003Cspan style=\"color:#B392F0\"> rate_limit_middleware\u003C/span>\u003Cspan style=\"color:#E1E4E8\">(request: Request, call_next):\u003C/span>\u003C/span>\n\u003Cspan class=\"line\">\u003Cspan style=\"color:#E1E4E8\">    client_ip \u003C/span>\u003Cspan style=\"color:#F97583\">=\u003C/span>\u003Cspan style=\"color:#E1E4E8\"> request.client.host\u003C/span>\u003C/span>\n\u003Cspan class=\"line\">\u003Cspan style=\"color:#E1E4E8\">    key \u003C/span>\u003Cspan style=\"color:#F97583\">=\u003C/span>\u003Cspan style=\"color:#F97583\"> f\u003C/span>\u003Cspan style=\"color:#9ECBFF\">\"rate_limit:\u003C/span>\u003Cspan style=\"color:#79B8FF\">{\u003C/span>\u003Cspan style=\"color:#E1E4E8\">client_ip\u003C/span>\u003Cspan style=\"color:#79B8FF\">}\u003C/span>\u003Cspan style=\"color:#9ECBFF\">\"\u003C/span>\u003C/span>\n\u003Cspan class=\"line\">\u003C/span>\n\u003Cspan class=\"line\">\u003Cspan style=\"color:#6A737D\">    # Kapasitas 10 token, refill 2 token per detik\u003C/span>\u003C/span>\n\u003Cspan class=\"line\">\u003Cspan style=\"color:#E1E4E8\">    allowed \u003C/span>\u003Cspan style=\"color:#F97583\">=\u003C/span>\u003Cspan style=\"color:#E1E4E8\"> rate_limit_lua(\u003C/span>\u003Cspan style=\"color:#FFAB70\">keys\u003C/span>\u003Cspan style=\"color:#F97583\">=\u003C/span>\u003Cspan style=\"color:#E1E4E8\">[key], \u003C/span>\u003Cspan style=\"color:#FFAB70\">args\u003C/span>\u003Cspan style=\"color:#F97583\">=\u003C/span>\u003Cspan style=\"color:#E1E4E8\">[\u003C/span>\u003Cspan style=\"color:#79B8FF\">10\u003C/span>\u003Cspan style=\"color:#E1E4E8\">, \u003C/span>\u003Cspan style=\"color:#79B8FF\">int\u003C/span>\u003Cspan style=\"color:#E1E4E8\">(time.time()), \u003C/span>\u003Cspan style=\"color:#79B8FF\">2\u003C/span>\u003Cspan style=\"color:#E1E4E8\">])\u003C/span>\u003C/span>\n\u003Cspan class=\"line\">\u003C/span>\n\u003Cspan class=\"line\">\u003Cspan style=\"color:#F97583\">    if\u003C/span>\u003Cspan style=\"color:#F97583\"> not\u003C/span>\u003Cspan style=\"color:#E1E4E8\"> allowed:\u003C/span>\u003C/span>\n\u003Cspan class=\"line\">\u003Cspan style=\"color:#F97583\">        raise\u003C/span>\u003Cspan style=\"color:#E1E4E8\"> HTTPException(\u003C/span>\u003Cspan style=\"color:#FFAB70\">status_code\u003C/span>\u003Cspan style=\"color:#F97583\">=\u003C/span>\u003Cspan style=\"color:#79B8FF\">429\u003C/span>\u003Cspan style=\"color:#E1E4E8\">, \u003C/span>\u003Cspan style=\"color:#FFAB70\">detail\u003C/span>\u003Cspan style=\"color:#F97583\">=\u003C/span>\u003Cspan style=\"color:#9ECBFF\">\"Too Many Requests. Pelan-pelan ya!\"\u003C/span>\u003Cspan style=\"color:#E1E4E8\">)\u003C/span>\u003C/span>\n\u003Cspan class=\"line\">\u003C/span>\n\u003Cspan class=\"line\">\u003Cspan style=\"color:#F97583\">    return\u003C/span>\u003Cspan style=\"color:#F97583\"> await\u003C/span>\u003Cspan style=\"color:#E1E4E8\"> call_next(request)\u003C/span>\u003C/span>\u003C/code>\u003C/pre>\n  \u003C/div>\n\u003C/div>\u003Ch2 id=\"jadi-algoritma-mana-yang-cocok-buat-kalian\">Jadi, algoritma mana yang cocok buat kalian?\u003C/h2>\n\u003Cp>Pada akhirnya, nggak ada algoritma yang menang di semua situasi. Token Bucket juara buat API publik karena ramah \u003Cem>burst\u003C/em>. Leaky Bucket menang kalau kalian butuh aliran yang benar-benar rata. Fixed Window cuma cocok buat proteksi kasar yang nggak butuh presisi tinggi. Sliding Window Counter jadi kompromi paling seimbang buat kuota ketat per jam atau per hari.\u003C/p>\n\u003Cp>Yang paling penting, hitungan kuotanya disimpan di \u003Cem>shared store\u003C/em> dan diperiksa secara atomik. Soalnya \u003Cem>rate limiter\u003C/em> yang salah hitung gara-gara \u003Cem>race condition\u003C/em> itu kayak kasir yang ngitung duit sambil merem—kelihatan jalan, tapi saldonya udah kacau dari awal.\u003C/p>\n\u003Cp>Selamat memilih algoritma ria, dan semoga \u003Cem>bucket\u003C/em> kalian nggak pernah bocor pas jam sibuk! 👋\u003C/p>\n","\u003Cul>\n\u003Cli>Token Bucket paling toleran terhadap lonjakan sesaat, Leaky Bucket ngeluarin \u003Cem>request\u003C/em> dengan laju rata, Fixed Window paling gampang tapi rawan \u003Cem>boundary spike\u003C/em>, dan Sliding Window Counter paling seimbang antara akurasi dan hemat memori.\u003C/li>\n\u003Cli>Token Bucket jadi standar \u003Cem>de facto\u003C/em> di \u003Cem>API Gateway\u003C/em> modern kayak AWS API Gateway, Kong, dan NGINX karena ramah sama \u003Cem>burst capacity\u003C/em>.\u003C/li>\n\u003Cli>Boundary spike di Fixed Window bisa ngelolosin \u003Cem>request\u003C/em> sampai dua kali lipat kuota tepat di pergantian interval waktu.\u003C/li>\n\u003Cli>Sliding Window Counter butuh Redis Hashes dan Lua Script biar operasi cek-kuota tetap atomik dan bebas \u003Cem>race condition\u003C/em> di arsitektur \u003Cem>microservices\u003C/em>.\u003C/li>\n\u003C/ul>\n",[416,419,422,430],{"id":320,"title":321,"slug":322,"excerpt":323,"thumbnail_image":417,"tags":418},{"id":325},[327,328,329,330,331],{"id":43,"title":44,"slug":45,"excerpt":46,"thumbnail_image":420,"tags":421},{"id":49},[51,52,53,54,55,56,57],{"id":423,"title":424,"slug":425,"excerpt":426,"thumbnail_image":427,"tags":429},"44","Transactional Outbox Pattern: Solusi Konsistensi Database dan Message Broker Tanpa 2PC","transactional-outbox-pattern-konsistensi-database-message-broker","Pernah nggak kalian lihat pesanan udah masuk database, tapi email konfirmasinya nggak pernah kekirim? Itu gejala *dual-write problem*. Yuk, kita bedah Transactional Outbox Pattern biar data dan *event* kalian nggak lagi berselisih paham!",{"id":428},"e3c617fd-1486-4b82-b937-94fbec846c77",[51,120,56,131,119,52,132],{"id":431,"title":432,"slug":433,"excerpt":434,"thumbnail_image":435,"tags":437},"43","Dead Letter Queue (DLQ) & Retry Policy: Menyelamatkan Pesan Nyasar di RabbitMQ Tanpa Bikin Worker Macet","dead-letter-queue-retry-policy-rabbitmq-resilient-messaging","Pernah nggak sih, *worker* kalian nolak satu pesan terus-terusan sampai CPU melonjak 100%? Itu namanya *poison message loop*, jebakan klasik di RabbitMQ. Yuk, kita bahas cara pasang *retry policy* dan *dead letter queue* yang bener!",{"id":436},"6c0a7ead-4841-4073-b29e-df7ae0f2aa3c",[118,119,51,56,120,121,279],[439,443,446,449,452,455,458],{"text":440,"depth":441,"id":442},"Token bucket, si favorit yang fleksibel",2,"token-bucket-si-favorit-yang-fleksibel",{"text":444,"depth":441,"id":445},"Leaky bucket, yang doyan ngerapiin antrean","leaky-bucket-yang-doyan-ngerapiin-antrean",{"text":447,"depth":441,"id":448},"Fixed window counter, simpel tapi ada jebakannya","fixed-window-counter-simpel-tapi-ada-jebakannya",{"text":450,"depth":441,"id":451},"Sliding window counter, akurat tanpa boros memori","sliding-window-counter-akurat-tanpa-boros-memori",{"text":453,"depth":441,"id":454},"Kapan pilih yang mana?","kapan-pilih-yang-mana",{"text":456,"depth":441,"id":457},"Implementasi praktis: token bucket dengan Redis dan Lua","implementasi-praktis-token-bucket-dengan-redis-dan-lua",{"text":459,"depth":441,"id":460},"Jadi, algoritma mana yang cocok buat kalian?","jadi-algoritma-mana-yang-cocok-buat-kalian",1791045070994]