{"article":{"annotation":{"article_id":"nvidia-agent-rewrites-spark-udf-gpu-workload-october-2026","available_at":"2026-10-15T19:01:05.650499Z","geography_revision":"ef6ee0010dda5fe7b240fd514749b43e214c57da73222d2e40ec83b276dd0977","id":"8cfa57071ac1dc424d97b5aaf23d80657d3cc783937f62c6fa7be66b335f3d84","input_sha256":"e76bd440656b64c28da5fb6295b8118fe7a99bed1a9cad01ca957a84d50fd443","limitations":["Rule-derived subjects are candidates, not verified editorial labels or calibrated model probabilities.","Coverage is distinct from event location. Unknown country/city/coordinates remain absent; no automatic geocoding.","Article geometries keep their stated precision and methodology; reproducing them is not independent coordinate verification."],"method":"controlled-rules-v1","places":[],"rules_sha256":"fb5ad3c27c91cb840f7a0a05f7f64a5cb26ad0374409ff9457f277c7092e22b4","schema_version":1,"subjects":[{"concept_id":"medtop:13000000","confidence":null,"confidence_method":null,"evidence":[{"field":"en.title+en.summary","language":"en","note":"Subject suggested from the title or summary: technology","source_urls":["https://developer.nvidia.com/blog/rewrite-apache-spark-udfs-gpu-acceleration-ai-agents/"]}],"method":"controlled-rules-v1","rule_id":"technology","status":"candidate"}],"taxonomy_revision":"1f2db31d811fca07e99b16d402c10e2cc2965c9df1c7b41d9be348b40a24e7fb","taxonomy_version":"IPTC Media Topics 2026-Q2"},"api_url":"https://newsverum.com/api/v1/articles/nvidia-agent-rewrites-spark-udf-gpu-workload-october-2026?lang=fr","category":"technology","content":{"key_points":[],"paragraphs":["NVIDIA explique que les fonctions définies par l’utilisateur (UDF) dans Apache Spark peuvent forcer un pipeline GPU à revenir à une exécution CPU, ajoutant des coûts de transfert de données et de calcul. Ses compétences cuDF guident les agents de programmation pour tester le comportement d’une UDF, la réécrire en Spark SQL, cuDF ou CUDA personnalisé, puis mesurer l’équivalence et les performances. L’article cite une entreprise technologique grand public non nommée qui a converti plus de 30 UDF en production grâce à ce processus.","Pour une charge de 600 milliards de lignes, le temps opérateur rapporté est de 274,55 heures avec une UDF sur CPU et de 12,78 heures avec une version cuDF sur GPU ; NVIDIA qualifie cela d’accélération par 22. L’article indique 16 To en entrée et 17,4 To en sortie, mais ne nomme pas le client et ne fournit pas de mesure indépendante. Ce résultat concerne une seule charge et ne garantit pas les performances de toutes les tâches Spark."],"sections":[]},"corrections":[],"country":null,"data":{"blocks":[{"description":{"ar":"أزمنة التنفيذ المعلنة باستخدام CPU وGPU لعبء عمل واحد من 600 مليار صف لدى شركة تقنية استهلاكية لم يُذكر اسمها.","de":"Gemeldete Laufzeiten der CPU- und GPU-Implementierung für eine Arbeitslast mit 600 Milliarden Zeilen bei einem nicht genannten Technologieunternehmen für Verbraucher.","en":"Reported CPU and GPU implementation times for one 600-billion-row workload at an unnamed consumer technology company.","es":"Tiempos comunicados para las versiones de CPU y GPU de una carga de 600.000 millones de filas en una empresa tecnológica de consumo no identificada.","fr":"Durées rapportées pour les versions CPU et GPU d’une charge de 600 milliards de lignes chez une entreprise technologique grand public non nommée.","pt":"Tempos divulgados para as versões CPU e GPU de uma carga de 600 mil milhões de linhas numa empresa de tecnologia de consumo não identificada.","tr":"Bir anonim tüketici teknoloji şirketinin 600 milyar satırlı iş yükünde bildirilen CPU ve GPU uygulama süreleri."},"id":"spark-udf-operator-time","items":[{"basis":"observed","label":{"ar":"زمن تنفيذ UDF باستخدام CPU","de":"Operatorlaufzeit der CPU-basierten UDF","en":"CPU-based UDF operator time","es":"Tiempo de operador de la UDF con CPU","fr":"Temps opérateur de l’UDF sur CPU","pt":"Tempo do operador da UDF com CPU","tr":"CPU tabanlı UDF operatör süresi"},"measure":{"aggregation":"unspecified","frequency":"irregular","geography":"world","id":"nvidia-spark-udf-operator-time-cpu","periodRole":"announcement-date","unitCode":"hour"},"period":"2026-10-13","unit":"hours","value":274.55},{"basis":"observed","label":{"ar":"زمن تنفيذ UDF المعجل بوحدة GPU عبر cuDF","de":"Operatorlaufzeit der GPU-beschleunigten cuDF-UDF","en":"cuDF GPU-accelerated UDF operator time","es":"Tiempo de operador de la UDF acelerada por GPU con cuDF","fr":"Temps opérateur de l’UDF accélérée par GPU avec cuDF","pt":"Tempo do operador da UDF acelerada por GPU com cuDF","tr":"cuDF ile GPU hızlandırmalı UDF operatör süresi"},"measure":{"aggregation":"unspecified","frequency":"irregular","geography":"world","id":"nvidia-spark-udf-operator-time-gpu","periodRole":"announcement-date","unitCode":"hour"},"period":"2026-10-13","unit":"hours","value":12.78}],"methodology":{"ar":"أزمنة تنفيذ مشغل CPU وcuDF/RapidsUDF لدالة إنتاج واحدة، وفق منشور NVIDIA بتاريخ 13 أكتوبر 2026. يصف المصدر عبء عمل من 600 مليار صف و16 تيرابايت من المدخلات؛ وهذا مثال واحد لا يُعمم على جميع مهام Spark.","de":"CPU- und cuDF/RapidsUDF-Operatorlaufzeiten für eine produktive UDF laut NVIDIA-Beitrag vom 13. Oktober 2026. Die Quelle beschreibt eine Arbeitslast mit 600 Milliarden Zeilen und 16 TB Eingabe; dies ist ein einzelnes Beispiel und lässt sich nicht auf alle Spark-Aufträge übertragen.","en":"CPU and cuDF/RapidsUDF operator times for one production UDF, as reported in NVIDIA's 13 October 2026 post. The source describes a 600-billion-row workload with 16 TB of input; this is one example and should not be generalized to all Spark jobs.","es":"Tiempos de operador de CPU y cuDF/RapidsUDF para una UDF de producción, según la publicación de NVIDIA del 13 de octubre de 2026. La fuente describe una carga de 600.000 millones de filas y 16 TB de entrada; es un ejemplo y no debe generalizarse a todos los trabajos de Spark.","fr":"Temps opérateur CPU et cuDF/RapidsUDF pour une UDF de production, selon l’article de NVIDIA du 13 octobre 2026. La source décrit une charge de 600 milliards de lignes et 16 To d’entrée ; il s’agit d’un exemple unique, non généralisable à toutes les tâches Spark.","pt":"Tempos do operador CPU e cuDF/RapidsUDF para uma UDF de produção, segundo a publicação da NVIDIA de 13 de outubro de 2026. A fonte descreve uma carga de 600 mil milhões de linhas e 16 TB de entrada; é um exemplo e não deve ser generalizado a todos os trabalhos Spark.","tr":"NVIDIA'nın 13 Ekim 2026 tarihli yazısında tek bir üretim UDF'si için verilen CPU ve cuDF/RapidsUDF operatör süreleri. Kaynak iş yükünü 600 milyar satır ve 16 TB girdi olarak tanımlar; sonuç tek bir örnektir, tüm Spark işlerine genellenemez."},"sourceUrls":["https://developer.nvidia.com/blog/rewrite-apache-spark-udfs-gpu-acceleration-ai-agents/"],"title":{"ar":"زمن تنفيذ دالة UDF المعاد كتابتها في Spark","de":"Operatorlaufzeit einer umgeschriebenen Spark-UDF","en":"Operator time for a rewritten Spark UDF","es":"Tiempo de operador de una UDF de Spark reescrita","fr":"Temps opérateur d’une UDF Spark réécrite","pt":"Tempo de operador de uma UDF Spark reescrita","tr":"Spark UDF yeniden yazımında işlemci süresi"},"type":"metrics","updatedAt":"2026-10-15T18:50:44+00:00"}],"version":1},"data_downloads":[{"formats":{"csv":"https://newsverum.com/api/v1/articles/nvidia-agent-rewrites-spark-udf-gpu-workload-october-2026/data/spark-udf-operator-time?lang=fr&format=csv","json":"https://newsverum.com/api/v1/articles/nvidia-agent-rewrites-spark-udf-gpu-workload-october-2026/data/spark-udf-operator-time?lang=fr&format=json"},"id":"spark-udf-operator-time"}],"data_types":["metrics"],"editorial_illustration":{"alt":"Illustration conceptuelle de la technologie, des processeurs et des informations connectées.","asset_id":"newsverum-v1-technology","avif_variants":[{"height":320,"url":"https://newsverum.com/assets/newsverum/illustrations/v1/technology-480.avif","width":480},{"height":480,"url":"https://newsverum.com/assets/newsverum/illustrations/v1/technology-720.avif","width":720},{"height":640,"url":"https://newsverum.com/assets/newsverum/illustrations/v1/technology-960.avif","width":960},{"height":800,"url":"https://newsverum.com/assets/newsverum/illustrations/v1/technology-1200.avif","width":1200},{"height":1024,"url":"https://newsverum.com/assets/newsverum/illustrations/v1/technology-1536.avif","width":1536}],"credit":"Newsverum","documentary":false,"height":1024,"kind":"symbolic-editorial-illustration","language":"fr","scope":"topic","url":"https://newsverum.com/assets/newsverum/illustrations/v1/technology-1536.webp","variants":[{"height":320,"url":"https://newsverum.com/assets/newsverum/illustrations/v1/technology-480.webp","width":480},{"height":640,"url":"https://newsverum.com/assets/newsverum/illustrations/v1/technology-960.webp","width":960},{"height":1024,"url":"https://newsverum.com/assets/newsverum/illustrations/v1/technology-1536.webp","width":1536}],"width":1536},"event_id":"event-fb5faa0b-ef92-5216-af7c-ce89b3eaa64b","event_period":"2026-10-13","event_period_basis":"event-or-reference-period","id":"nvidia-agent-rewrites-spark-udf-gpu-workload-october-2026","image":null,"importance":0.6,"language":"fr","modified_at":"2026-10-15T19:00:45+00:00","publication_basis":"publication-record","published_at":"2026-10-15T19:00:45+00:00","reading_minutes":1,"region":"world","section_illustration":{"alt":"Illustration conceptuelle de la technologie, des processeurs et des informations connectées.","asset_id":"newsverum-v1-technology","avif_variants":[{"height":320,"url":"https://newsverum.com/assets/newsverum/illustrations/v1/technology-480.avif","width":480},{"height":480,"url":"https://newsverum.com/assets/newsverum/illustrations/v1/technology-720.avif","width":720},{"height":640,"url":"https://newsverum.com/assets/newsverum/illustrations/v1/technology-960.avif","width":960},{"height":800,"url":"https://newsverum.com/assets/newsverum/illustrations/v1/technology-1200.avif","width":1200},{"height":1024,"url":"https://newsverum.com/assets/newsverum/illustrations/v1/technology-1536.avif","width":1536}],"credit":"Newsverum","documentary":false,"height":1024,"kind":"symbolic-editorial-illustration","language":"fr","scope":"section","url":"https://newsverum.com/assets/newsverum/illustrations/v1/technology-1536.webp","variants":[{"height":320,"url":"https://newsverum.com/assets/newsverum/illustrations/v1/technology-480.webp","width":480},{"height":640,"url":"https://newsverum.com/assets/newsverum/illustrations/v1/technology-960.webp","width":960},{"height":1024,"url":"https://newsverum.com/assets/newsverum/illustrations/v1/technology-1536.webp","width":1536}],"width":1536},"slug":"nvidia-agent-rewrites-spark-udf-gpu-workload-october-2026","sources":[{"checkedAt":"2026-10-15","name":"NVIDIA Technical Blog","publishedAt":"2026-10-13","url":"https://developer.nvidia.com/blog/rewrite-apache-spark-udfs-gpu-acceleration-ai-agents/"}],"summary":"NVIDIA rapporte qu’une réécriture d’UDF Spark a réduit le temps opérateur de 274,55 à 12,78 heures sur une charge de 600 milliards de lignes chez une entreprise technologique grand public non nommée.","text":"NVIDIA rapporte qu’une réécriture d’UDF Spark a réduit le temps opérateur de 274,55 à 12,78 heures sur une charge de 600 milliards de lignes chez une entreprise technologique grand public non nommée.\n\nNVIDIA explique que les fonctions définies par l’utilisateur (UDF) dans Apache Spark peuvent forcer un pipeline GPU à revenir à une exécution CPU, ajoutant des coûts de transfert de données et de calcul. Ses compétences cuDF guident les agents de programmation pour tester le comportement d’une UDF, la réécrire en Spark SQL, cuDF ou CUDA personnalisé, puis mesurer l’équivalence et les performances. L’article cite une entreprise technologique grand public non nommée qui a converti plus de 30 UDF en production grâce à ce processus.\n\nPour une charge de 600 milliards de lignes, le temps opérateur rapporté est de 274,55 heures avec une UDF sur CPU et de 12,78 heures avec une version cuDF sur GPU ; NVIDIA qualifie cela d’accélération par 22. L’article indique 16 To en entrée et 17,4 To en sortie, mais ne nomme pas le client et ne fournit pas de mesure indépendante. Ce résultat concerne une seule charge et ne garantit pas les performances de toutes les tâches Spark.","title":"NVIDIA décrit la réécriture de fonctions Spark par des agents pour l’exécution GPU","topics":[],"translations":[{"api_url":"https://newsverum.com/api/v1/articles/nvidia-agent-rewrites-spark-udf-gpu-workload-october-2026?lang=tr","language":"tr","url":"https://newsverum.com/tr/news/nvidia-agent-rewrites-spark-udf-gpu-workload-october-2026"},{"api_url":"https://newsverum.com/api/v1/articles/nvidia-agent-rewrites-spark-udf-gpu-workload-october-2026?lang=en","language":"en","url":"https://newsverum.com/en/news/nvidia-agent-rewrites-spark-udf-gpu-workload-october-2026"},{"api_url":"https://newsverum.com/api/v1/articles/nvidia-agent-rewrites-spark-udf-gpu-workload-october-2026?lang=ar","language":"ar","url":"https://newsverum.com/ar/news/nvidia-agent-rewrites-spark-udf-gpu-workload-october-2026"},{"api_url":"https://newsverum.com/api/v1/articles/nvidia-agent-rewrites-spark-udf-gpu-workload-october-2026?lang=fr","language":"fr","url":"https://newsverum.com/fr/news/nvidia-agent-rewrites-spark-udf-gpu-workload-october-2026"},{"api_url":"https://newsverum.com/api/v1/articles/nvidia-agent-rewrites-spark-udf-gpu-workload-october-2026?lang=es","language":"es","url":"https://newsverum.com/es/news/nvidia-agent-rewrites-spark-udf-gpu-workload-october-2026"},{"api_url":"https://newsverum.com/api/v1/articles/nvidia-agent-rewrites-spark-udf-gpu-workload-october-2026?lang=de","language":"de","url":"https://newsverum.com/de/news/nvidia-agent-rewrites-spark-udf-gpu-workload-october-2026"},{"api_url":"https://newsverum.com/api/v1/articles/nvidia-agent-rewrites-spark-udf-gpu-workload-october-2026?lang=pt","language":"pt","url":"https://newsverum.com/pt/news/nvidia-agent-rewrites-spark-udf-gpu-workload-october-2026"}],"url":"https://newsverum.com/fr/news/nvidia-agent-rewrites-spark-udf-gpu-workload-october-2026"},"version":"1"}