Emerging Embdedded Memories and Storage
summary
Session presents four implemented systems with quantifiable gains: SPF adaptive synchronous handler changes OS scheduling plus selective prefetching and saves about 12% (SPFstore) and 9% (SPF) total execution time on a trace-driven simulator by reducing synchronous page-fault peaks and working-set churn; SkyTree skyrmion-friendly B+ tree redesigns node layout and recycling to cut search latency ~73% and energy 67–73% in racetrack-memory simulation; adaptive active-block allocation plus partial refresh shortens 3D‑NAND open times and lowers page-copy/GC overheads; WA-OPShare workload-adaptive OP allocation reallocates over-provisioning to boost bandwidth up to ~40% and reduce write amplification ~37% in multi-tenant SSD simulation.
AI-written summary · about
chapters (6)
- 0:00Emerging embedded memories
- 2:24Page fault handling challenges
- 22:10B+3 meets racetrack indexing
- 39:543D TLC NAND reliability
- 52:50SSD over-provisioning for workloads
- 1:04:37Q&A
related talks
15mImproving Performance of Flash Based Key-Value Stores Using Storage Class Memory...
27mPreview: Storage
43mPromise and Pitfalls of Persistent Memory
17mPreview: Persistent Memory
35mDeveloping Software for Persistent Memory
06mEarly experience of Utilizing Persistent Memory for Database... (Work-in-Progress Report)
35mDeveloping Software for Persistent Memory
19mPolyStore: Exploiting Combined Capabilities of Heterogeneous Storage
