MAPX: Controlled Data Migration in the Expansion of Decentralized Object Based Storage
summary
MAPX extends CRUSH by adding a virtual layer beneath root for each expansion and controlling Placement Group placement via static and dynamic PG timestamps, avoiding bulk migration. It implements a per-object timestamp in RBD header so new objects map to new layers; PGs remap by changing TPG_d. Prototype on Ceph-RBD shows up to 4.25× better 99th‑percentile latency and ~74% lower tail latency in some tests, mapping costs only tens of microseconds, and scales across hundreds to thousands of OSDs.
AI-written summary · about
chapters (5)
related talks
02h 59mCeph Storage: a Caffeinated Primer
49mIntro to Ceph, the Distributed Storage System
17mDuhu: Shared Disaggregated Memory for Distributed Data Processing Frameworks
30mScaling Databases and File APIs with Programmable Ceph Object Storage
06mNobody Knows What PGs are Good For, Only I Do
26mImplementing a Resilient Ceph RGW Object Storage with Erasure Coded Data Pool as the Foundation...
45mBuilding Available and Partition Tolerant...
45mBeyond Machine Boundaries: Experience with Memory Disaggregation