Early experience of Utilizing Persistent Memory for Database... (Work-in-Progress Report)
summary
Compares using persistent memory vs DRAM for bulk loading TPCH lineitem on a homegrown database engine, measuring four configs (DRAM/PMem as sort buffer and intermediate space). Finds 2.2× slower single-threaded overall when persistent memory replaces DRAM for sort buffering, with the gap shrinking to 1.3× at 96 threads and the sort step itself up to 3.8× slower single-threaded; merging (sequential reads) shows comparable speed. Proposes a static/dynamic allocation optimizer for heterogeneous memory.
AI-written summary · about
chapters (6)
related talks
35mDeveloping Software for Persistent Memory
17mPreview: Persistent Memory
35mDeveloping Software for Persistent Memory
43mPromise and Pitfalls of Persistent Memory
52mByte-Addressable Persistent Memory and DBMS
15mImproving Performance of Flash Based Key-Value Stores Using Storage Class Memory...
16mJohnny Cache: the End of DRAM Cache Conflicts (in Tiered Main Memory Systems)
01h 07mEmerging Embdedded Memories and Storage