# Storage I/O Engineering: Dataset'in Eğitimi Yavaşlatma Sanatı (ve Önleme)

> Source: https://sukruyusufkaya.com/learn/fine-tuning-cookbook/ftc-storage-io-engineering-dataset-bottleneck
> Updated: 2026-08-17T03:18:46.470Z
> Category: Fine-Tuning Cookbook (Model-by-Model)
> Module: Part I — Hardware & Memory Engineering
**TLDR:** Dataset bottleneck: GPU %30 idle bekliyor çünkü disk yetmiyor. NVMe Gen3/Gen4/Gen5 throughput, dataset format seçimi (parquet vs arrow vs webdataset), HuggingFace datasets caching, num_workers tuning, prefetch_factor, persistent_workers, pinned memory, FSx vs S3 vs local — RTX 4090 + 50K Türkçe dataset'i 0 idle çalıştırma reçetesi.