User Tools

Site Tools


internal:projects:hybrid-memory:start

Table of Contents

Hybrid Memory

Low-energy and low-cost hybrid memory systems built out of small amount of RAM and large flash-based storage.

Participants

Meetings

  • 02/25/11: Implementation update
  • 02/03/11: Implementation update
  • 12/02/10: Update on Linux HM implementation
  • 11/10/10: More updates and alternative approaches to simulation optimizations
  • 11/04/10: Updates on performance troubleshooting and New direction on 'store faults'?
  • 11/02/10: Tracing hardware events initial implementation discussion and alternatives for dealing with high page fault I/O latencies in design (buy a Fusion-IO or OCZ card?)
  • 10/21/10: Reviewing new method for memory traces, Alternatives for power computation
  • 10/07/10: New approach for memory traces
  • 09/30/10: Validation of QEMU traces using the follow benchmark and power calculations
  • 09/23/10: Pre-eurosys meeting
  • 09/16/10: Start of Eurosys submission discussion
  • 05/06/10: Another Pre-OSDI meeting
  • 05/04/10: Pre-OSDI meeting
  • 05/03/10: Next tasks for Motivation and Evaluation sections
  • 04/29/10: Debugging power numbers and memory bandwidth, and moving to flash power model
  • 04/26/10: Motivation graphs and writing tasks and evaluation section updates.
  • 04/22/10: Evaluation section - first round of graphs for latency, erasures.
  • 04/19/10: Latency numbers and power computation issues
  • 04/15/10: Brief discussion of power computations and writing plan
  • 04/05/10: New results with SSD size = 4x mem-footprint and evaluation section plans
  • 04/02/10: Evaluation section and QEMU trace timing bug
  • 03/29/10: Plans for new experiments, SpecPower benchmark addition, and Stable state discussion
  • 03/15/10: Discussion of cold bandwidth and spreadsheet data with Akshat
  • 03/11/10: New spreadsheet data + Simulation accuracy and timing of replay modes
  • 03/08/10: Latency CDFs for first round of experiments
  • 02/18/10: Cold bandwidth numbers discussion
  • 02/02/10: Synchronization issues with HM implementation
  • 01/26/10: Motivation graph formats and Padding discussion
  • 01/15/10: Plan for OSDI submission
  • 12/10/09: Cost-based replacement scrapped - looking at alternatives: CFLRU, ARC, and CFARC
  • 12/03/09: Page-fault w/ cost-based replacement + bandwidth for cold data redone
  • 11/25/09: Miscellaneous discussion
  • 11/19/09: Memory trace processing, new mobile traces, emulator bug
  • 11/12/09: Resolving emulator inefficiencies and power model data needed
  • 11/09/09: Flushing timer and measuring memory latencies
  • 11/05/09: Timing accurate replay in emulator
  • 11/02/09: Graphs for motivation
  • 10/29/09: Replacement algorithm revision, phases and other motivation data
  • 10/26/09: Motivation beyond flashcache - separation of anonymous and page-cache references
  • 10/15/09: QEMU tracing optimizations, prototype design issues, power models, etc.
  • 10/08/09: Update on Prototype, QEMU tracing, and MICRON memory power models discussion
  • 10/02/09: Performance counters Power6 and Opteron for memory traces
  • 10/01/09: Luis comments for paper
  • 09/03/09: More PTLSim issues
  • 08/31/09: Simulator/Emulator design issues and PTLSim updates
  • 08/24/09: Paper comments
  • 08/20/09: Paper comments and writing plan
  • 08/17/09: Simulator overview discussion, PTLSim issues
  • 08/06/09: Draft comments from Luis
  • 08/03/09: PTLSim slowdown, steady-state-only simulation, emulation setup ideas
  • 07/30/09: Using page-reference bits, motivation graphs
  • 07/27/09: Discussion of benchmarks and motivation graph 1
  • 07/23/09: Discussion of motivation graphs and effect of missing DMA
  • 07/20/09: Paper restructuring and taxonomy of design alternatives
  • 07/16/09: Details of cost-benefit approach for cleaning and padding
  • 07/13/09: Comments on segment selection (change of terminology)
  • 07/09/09: New thoughts on locality-aware flushing and refining segment selection
  • 07/06/09: PTLSim updates, ideas on segment selection
  • 07/02/09: Page Replacement candidate formalization
  • 06/29/09: PLTSim updates, Paper feedback from Luis
  • 06/25/09: Brief design discussion and Old QEMU success
  • 06/22/09: Paper outline, brief design, and identifying simulator
  • 06/18/09: Possibilities for memory access traces
  • 06/15/09: Requirements of development / advantages with QEMU and OSes other than Linux
  • 06/11/09: Discussion of evaluation methods, alternate (MTD-based) architecture/devices, allocation and replacement for VM
internal/projects/hybrid-memory/start.txt · Last modified: 2024/06/28 20:42 by 127.0.0.1