OSCA: An Online-Model Based Cache Allocation Scheme in Cloud Block Storage Systems
summary
OSCA presents an online cache allocator that builds miss-ratio curves from a low-overhead re-access ratio O(1) model (no trace collection), converts those MRCs into per-node cache requirements, and searches for near-optimal sizes with a dynamic programming allocator. Implemented as a cache controller on real-world CBS traces, it achieves MAE comparable to state-of-the-art sampling methods, runs daily profiling, and yields a measured 13.2% backend IO reduction versus equal allocation.
AI-written summary · about
chapters (16)
- 0:00Intro
- 1:54Background
- 3:11Motivation
- 7:03Main Ideas
- 8:16Cache Modeling (cont.)
- 9:48Reuse Distance
- 12:02Obtain Re-access Ratio
- 12:23Re-access Ratio Curve
- 13:53Construct MRC from RAR
- 14:46Define the Optimization Target
- 15:18Search for the Optimal Solution
- 15:41Dynamical Programming
- 16:22System Evaluations • Trace Collection
- 17:49Miss Ratio Curves
- 18:37Mean Absolute Error (MAE)
- 19:00Overall Efficacy
related talks
11meMRC: Efficient Miss Ratio Approximation for Multi-Tier Caching
20mAdaptive Online Cache Capacity Optimization via Lightweight Working Set Size...
17mScalaCache: Scalable User-Space Page Cache Management with Software-Hardware...
19mKosmo: Efficient Online Miss Ratio Curve Generation for Eviction Policy Evaluation
20mSymbiosis: The Art of Application and Kernel Cache Cooperation
25mDistCache: Provable Load Balancing for Large-Scale Storage Systems with Distributed...
19mA zone-aware cache system for distributed databases
17mPushing the Limits of In-Network Caching for Key-Value Stores