作者: Hans-Arno Jacobsen , Tilmann Rabl , Mohammad Sadoghi , Prashanth Menon
DOI:
关键词: Value (computer science) 、 Embedded system 、 Computer science 、 Throughput (business) 、 Big data 、 Key (cryptography) 、 Latency (engineering) 、 Hybrid storage 、 Data access
摘要: Flash-based solid state drives (SSDs) are increasingly becoming a popular choice as storage device within database management systems and key-value stores alike. SSDs offer fast throughput low latency access to data, but their price-per-byte cost often makes them uneconomical for exclusive use, especially in the era of big data workloads. A common solution this problem is augment existing by adding smaller that target only performance-critical areas. We believe hybrid approach be stop-gap solution. Rather than simply extending with SSDs, work we completely re-architect how operates setting both small slower high-capacity HDDs. formulate an accurate I/O model study behave under several varying representative Based on these studies taking holistic approach, design system dynamically optimizes layout strategy leverages strengths each available medium.