Bimodal Caching

Participants

Steven Lyons
Raju Rangaswami
Jason Liu
Leonardo Marmol
Ricardo Santana

Project Goals

  Workload stability is not addressed in current caching policies. We need a new one.

Results

Summary Graphs

Summary of HM Traces

Summary of MDS Traces

Summary of PRN Traces

Summary of PROJ Traces

Summary of PRXY Traces

Summary of RSRCH Traces

Summary of SRC1 Traces

Summary of SRC2 Traces

Summary of STG Traces

Summary of TS Trace

Summary of USR Traces

Summary of WDEV Traces

Summary of WEB Traces

Ghost LRU

Frequency Distribution

Graphs

ARC Event Visualization

Graphs

ARC and LARC Miss Visualization

Miss Graphs

Meetings

07/02/15: New plots; Updates on kernel module
06/30/15: Analysis discussion; new write-rate data
06/26/15: Updates on kernel module; Multi-level cache partitioning
06/23/15: Initial MARC-Adaptive numbers for FIU and MSR
06/19/15: Webusers deep-dive + module update
06/16/15: MSR OPT and OPT-F data
06/16/15: TBD
06/12/15: A-U plots, motivation
06/09/15: Updates on OPT results and ARC analysis brainstorming
06/05/15: (Evening) Paper discussion (intro/motivation)
06/05/15: Sensitivity data for sample size with mARC
06/02/15: Algorithm discussion
05/22/15: Module, analysis, and OPT discussion
05/15/15: Adaptive mARC – new ideas
05/12/15: Brainstorming Adaptive with Giri
05/08/15: HotStorage final version plans (Ricardo delivers on Tuesday) + clarification of Oracle
05/05/15: HotStorage Final version plans and Discussion of Oracle Algorithm
04/24/15: HotStorage final version paper improvement plan
04/17/15: Making m-ARC fully adaptive!
04/14/15: Updates on m-ARC windowing and module
04/07/15: Updates on module
04/03/15: Updates on module, testbed, and m-ARC improvement plans
03/31/15: Full paper plans (storyline); kernel module status in development
03/27/15: Algorithm change plans and module implementation updates
03/24/15: ACM SOCC - submission plans and task list
03/20/15: ACM SOCC - submission plans and task list
03/16/15: HotStorage15 submission final tasks
03/06/15: HotStorage15 paper status and plans
03/05/15: HotStorage15 paper status and plans
03/03/15: HotStorage15 paper status and plans
02/27/15: HotStorage15 paper status and plans
02/24/15: TBD
02/13/15: TBD
02/10/15: TBD
02/06/15: TBD
01/27/15: TBD
01/23/15: New latency data discussion
01/20/15: Latency data
01/16/15: General updates
01/08/15: Updates and next steps for paper writing / data generation
12/19/14: Steven discussion of new ideas
12/16/14: Paper writing tasks
12/12/14: Cost model discussion
12/09/14: Santana overview (Koller joins in)
12/09/14: Updates from Steven on the stats work and plots
12/02/14: Ricardo new plots; Steven/Andy on trace processing
11/14/14: Ricardo/Andy updates; Asana project management overview discussion
11/06/14: Paper outline (partial review) and writing plans
10/31/14: Ricardo's paper storyline and writing tasks
10/21/14: Steven updates
10/17/14: Andy Martinez introduction; Steven updates
10/10/14: Ricardo updates
10/03/14: Revisiting rules for bimodal caching
09/30/14: Updates on new data from Ricardo
09/29/14: Tabulation of motivation rules for Bimodal Caching
09/16/14: Progress status
09/09/14: Ricardo's new algorithm
09/05/14: Workload modeling introduction by Jason
09/02/14: New algorithm discussion
08/28/14: Discussion with Jason on workload modeling
08/26/14: (Steven, Jason) Modeling workloads and stability; entry/exit conditions for stability
08/21/14: (Ricardo) Outcome of miss/hit count variants not as good as original R-ARC
08/19/14: (Steven, Ricardo) Reducing False positive detection of stability
08/19/14: (Ricardo) New data and plans
08/12/14: Stability metrics? Analyzing ARC events
08/08/14: Rank-based bimodal caching (brainstorming)
08/05/14: Motivation section (observations) (whiteboard)
07/29/14: Reuse stability within a phase (data)
07/22/14: Motivating stability; phase detection techniques; writes vs. reads; etc.
07/15/14: New data for motivating bimodal arc
07/08/14: Bimodal arc data collection plans
07/01/14: New graphs and new directions with dynamic granularity and shadow ARC within Bimodal ARC
06/10/14: Examining warm-up data and initial review of paper draft; miss-rate seems buggy
06/06/14: Evaluating variants (always populate if space vs. not); next step: warming up caches
05/27/14: Initial ARC data – how about enabling population if we have cache space?
04/29/14: (Steven) Paper storyine brainstorming
04/24/14: (Ricardo) New algorithms and paper outline!
04/22/14: Brainstorming and Recap of new algorithms
04/15/14: New data and algorithm
04/08/14: Data for miss/hit sequence lengths, sensitivity to fixed threshold
04/01/14: More data analysis and new algorithm?
03/25/14: Brainstorming and results from 15% fixed threshold; next step is to explore stability based trigger
03/18/14: Brainstorming ideas for working-set and cache usage modeling
03/18/14: Initial results with coarse-grained threshold-based on/off for caching; plans for new experiments
03/11/14: Revisiting caching algorithm and new miss-rate data (multi-day)
03/04/14: 1GB = 16GB for miss-rates? Modeling working-sets and other things
02/25/14: 4GB caches with individual workloads is still high miss-rate for most; how does background I/O activity affect phase detection (e.g., pdflush and kjournald)?
02/11/14: Discussion of write-policies; Cache sizes for individual traces needs to be increased
02/04/14: Analyzing individual workloads instead of combined
01/30/14: New averaging discussion
01/28/14: Back to basics
01/21/14: Filtering derivative peaks and possible next steps
01/14/14: What are we missing in detecting phases?
01/07/14: More miss-rate histograms; detecting phases with derivative plots
01/03/14: Analyzing miss-rate histograms; defining averaging granularities for detecting phases
12/23/13: What is phase stability?
12/13/13: Updates on using new LRU simulator code, tuning cache size, averaging window size
12/04/13: Project goals revisited and bug finding
11/15/13: Project goal and current direction

Systems Research Laboratory

Table of Contents

Bimodal Caching

Participants

Project Goals

Results

Homes Trace

Online and Web Traces

Working Set

Cambridge Traces

Summary Graphs

Ghost LRU

Frequency Distribution

ARC Event Visualization

ARC and LARC Miss Visualization

Meetings