Designing Floating Codes for Expected Performance Hilary Finucane Zhenming Liu Michael Mitzenmacher.

Slides:



Advertisements
Similar presentations
Why Simple Hash Functions Work : Exploiting the Entropy in a Data Stream Michael Mitzenmacher Salil Vadhan And improvements with Kai-Min Chung.
Advertisements

Introduction to Flash Memories And Flash Translation Layer
Ulams Game and Universal Communications Using Feedback Ofer Shayevitz June 2006.
I/O Management and Disk Scheduling Chapter 11. I/O Driver OS module which controls an I/O device hides the device specifics from the above layers in the.
The Capacity of Wireless Networks
On-line learning and Boosting
Why Simple Hash Functions Work : Exploiting the Entropy in a Data Stream Michael Mitzenmacher Salil Vadhan.
©Silberschatz, Korth and Sudarshan12.1Database System Concepts Chapter 12: Part C Part A:  Index Definition in SQL  Ordered Indices  Index Sequential.
©Silberschatz, Korth and Sudarshan12.1Database System Concepts Chapter 12: Indexing and Hashing Basic Concepts Ordered Indices B+-Tree Index Files B-Tree.
Bounds on Code Length Theorem: Let l ∗ 1, l ∗ 2,..., l ∗ m be optimal codeword lengths for a source distribution p and a D-ary alphabet, and let L ∗ be.
CS 261- Winter 2009 Dynamic Array Queue and Deque.
Allocating Memory.
Model Assessment, Selection and Averaging
Cuckoo Hashing : Hardware Implementations Adam Kirsch Michael Mitzenmacher.
Some Results on Codes for Flash Memory Michael Mitzenmacher Includes work with Hilary Finucane, Zhenming Liu, Flavio Chierichetti.
June 3, 2015Windows Scheduling Problems for Broadcast System 1 Amotz Bar-Noy, and Richard E. Ladner Presented by Qiaosheng Shi.
Kuang-Hao Liu et al Presented by Xin Che 11/18/09.
Clustering short time series gene expression data Jason Ernst, Gerard J. Nau and Ziv Bar-Joseph BIOINFORMATICS, vol
On the Topologies Formed by Selfish Peers Thomas Moscibroda Stefan Schmid Roger Wattenhofer IPTPS 2006 Santa Barbara, California, USA.
1 Data Persistence in Large-scale Sensor Networks with Decentralized Fountain Codes Yunfeng Lin, Ben Liang, Baochun Li INFOCOM 2007.
Tirgul 10 Rehearsal about Universal Hashing Solving two problems from theoretical exercises: –T2 q. 1 –T3 q. 2.
Codes for Deletion and Insertion Channels with Segmented Errors Zhenming Liu Michael Mitzenmacher Harvard University, School of Engineering and Applied.
Why Simple Hash Functions Work : Exploiting the Entropy in a Data Stream Michael Mitzenmacher Salil Vadhan.
Lecture 5: Learning models using EM
1 On Compressing Web Graphs Michael Mitzenmacher, Harvard Micah Adler, Univ. of Massachusetts.
Tirgul 8 Universal Hashing Remarks on Programming Exercise 1 Solution to question 2 in theoretical homework 2.
Santa Clara, CA USA August An Information Theory Approach for Flash Memory Eitan Yaakobi, Paul H. Siegel, Jack K. Wolf University of California,
Optimal Fast Hashing Yossi Kanizo (Technion, Israel) Joint work with Isaac Keslassy (Technion, Israel) and David Hay (Politecnico di Torino, Italy)
Coding for Flash Memories
1 Verification Codes Michael Luby, Digital Fountain, Inc. Michael Mitzenmacher Harvard University and Digital Fountain, Inc.
CSI Uncertainty in A.I. Lecture 201 Basic Information Theory Review Measuring the uncertainty of an event Measuring the uncertainty in a probability.
. Expressive Graphical Models in Variational Approximations: Chain-Graphs and Hidden Variables Tal El-Hay & Nir Friedman School of Computer Science & Engineering.
Monte Carlo Methods in Partial Differential Equations.
1 Introduction to Quantum Information Processing QIC 710 / CS 768 / PH 767 / CO 681 / AM 871 Richard Cleve QNC 3129 Lecture 18 (2014)
Asaf Cohen (joint work with Rami Atar) Department of Mathematics University of Michigan Financial Mathematics Seminar University of Michigan March 11,
Frame by Frame Bit Allocation for Motion-Compensated Video Michael Ringenburg May 9, 2003.
1 Introduction to Quantum Information Processing QIC 710 / CS 667 / PH 767 / CO 681 / AM 871 Richard Cleve DC 2117 Lecture 16 (2011)
: Chapter 12: Image Compression 1 Montri Karnjanadecha ac.th/~montri Image Processing.
« Performance of Compressed Inverted List Caching in Search Engines » Proceedings of the International World Wide Web Conference Commitee, Beijing 2008)
CSC 41/513: Intro to Algorithms Linear-Time Sorting Algorithms.
Amortized Analysis Typically, most data structures provide absolute guarantees on the worst case time for performing a single operation. We will study.
Markov Decision Processes1 Definitions; Stationary policies; Value improvement algorithm, Policy improvement algorithm, and linear programming for discounted.
Quality of Service Karrie Karahalios Spring 2007.
1 ECE-517 Reinforcement Learning in Artificial Intelligence Lecture 7: Finite Horizon MDPs, Dynamic Programming Dr. Itamar Arel College of Engineering.
Learning Lateral Connections between Hidden Units Geoffrey Hinton University of Toronto in collaboration with Kejie Bao University of Toronto.
The Selection Problem. 2 Median and Order Statistics In this section, we will study algorithms for finding the i th smallest element in a set of n elements.
David Luebke 1 10/25/2015 CS 332: Algorithms Skip Lists Hash Tables.
Project 1: DRAM timing violation due to PV Due to PV, transistor and capacitor may have variations in their dimensions, causing charging time of a cell.
Coding and Algorithms for Memories Lecture 5 1.
Heuristic Optimization Methods Greedy algorithms, Approximation algorithms, and GRASP.
Error Correction and Partial Information Rewriting for Flash Memories Yue Li joint work with Anxiao (Andrew) Jiang and Jehoshua Bruck.
Coding and Algorithms for Memories Lecture 4 1.
CSC401 – Analysis of Algorithms Lecture Notes 2 Asymptotic Analysis Objectives: Mathematics foundation for algorithm analysis Amortization analysis techniques.
Similarity Searching in High Dimensions via Hashing Paper by: Aristides Gionis, Poitr Indyk, Rajeev Motwani.
The famous “sprinkler” example (J. Pearl, Probabilistic Reasoning in Intelligent Systems, 1988)
De novo discovery of mutated driver pathways in cancer Discussion leader: Matthew Bernstein Scribe: Kun-Chieh Wang Computational Network Biology BMI 826/Computer.
1 Use graphs and not pure logic Variables represented by nodes and dependencies by edges. Common in our language: “threads of thoughts”, “lines of reasoning”,
Stochastic Optimization for Markov Modulated Networks with Application to Delay Constrained Wireless Scheduling Michael J. Neely University of Southern.
CSE280Stefano/Hossein Project: Primer design for cancer genomics.
Coding and Algorithms for Memories Lecture 6 1.
Unified Adaptivity Optimization of Clock and Logic Signals Shiyan Hu and Jiang Hu Dept of Electrical and Computer Engineering Texas A&M University.
Coding and Algorithms for Memories Lecture 2
Finite state machine optimization
Finite state machine optimization
Coding and Algorithms for Memories Lecture 5
Introduction to Algorithms
Randomized Algorithms CS648
Linear-Time Sorting Algorithms
The Selection Problem.
Bin Packing Michael T. Goodrich Some slides adapted from slides from
Presentation transcript:

Designing Floating Codes for Expected Performance Hilary Finucane Zhenming Liu Michael Mitzenmacher

Floating Codes A new model for flash memory. State is an n-ary sequence of q-ary numbers. –Represents block of n cells; each cell holds an electric charge. State mapped to variable values. –Gives k-ary sequence of l-ary numbers. State changes by increasing one or more cell values, or reset entire block. –Adding charge is easy; removing charge requires resetting everything. –Resets are expensive!!!!

Floating Codes: The Problem As variable values change, need state to track variables. How do we choose the mapping function from states to variables AND the transition function from variable changes to state changes to maximize the time between reset operations?

History Write-once memories (WOMs) –Rivest and Shamir, early 1980’s. –Punch cards, optimal disks. –Can turn 0’s to 1’s, but not back again. Many related models: WOMs, WAMs, WEMs, WUMs. Floating codes (Jiang, Bohossian, Bruck) use model for Flash Memory. –Designed to maximize worst-case time between resets.

Contribution : Expected Time Argument: Worst-case time between resets is not right design criterion. –Many resets in a lifetime. –Mass-produced product. –Potential to model user behavior. Statistical performance guarantees more appropriate. –Expected time between resets. –Time with high probability. –Given a model.

Specific Contributions Problem definition / model Codes for simple cases

Formal Model General Codes We consider limited variation; one variable changes per step.

Formal Model Continued Above : when –Cost is 0 when R moves to cell state above previous, 1 otherwise. Assumption : variables changes given by Markov chain. –Example : ith bit changes with prob. p i –Given D, R, gives Markov chain on cell states. –Let  be equilibrium on cell states. Goal is to minimize average cost: –Same as maximize average time between resets.

Variations Many possible variations –Multiple variables change per step –More general random processes for values –Rules limiting transitions –General costs, optimizations Hardness results? –Conjecture some variations NP-hard or worse.

Specific Case Binary values : l = 2 2 bits : k = 2 Markov model: only bit changes at each step. First bit changes with probability p. Result : Asympotically optimal code. –Code handles n(q-1)-o(nq) value changes with high probability. –Same code works for every value of p.

Code : n = 2, k = 2, l = 2 2 bit values. 2 cells. Code based on striped Gray code. Expected time/time with high probability before reset = 2q - o(q) Asymptotically optimal for all p, 0 < p < 1. Worst case optimal: approx 3q/2. D(0,0) = 00 D(1,3) = 11 R((1,0),2,1) = (2,0)

Proof Sketch “Even cells”: down with probability p, right with probability 1-p. “Odd cells” : right with probability p, down with probability 1-p. Code hugs the diagonal. Right/down moves approximately balance for first 2q-o(q) steps.

Performance Results Scheme DWC, DGC, DGC+, DWC, DGC, DGC+,

Code : n = 3, k = 2, l = 2 Layer Gray codes for n = 3. Expected time/time whp before reset = 3q - o(q) –Slightly hard argument. Asymptotically optimal for all p, 0 < p < 1.

Codes for k = l = 2 Glue together codes for larger n. Example : n = 4. Go 2q - o(q) in first two dimension, 2q - o(q) in next two, so 4q - o(q) overall. Some further results in paper.

Conclusions Introduce problem of maximizing expected time until a reset for floating codes. Simple schemes for k = 2, l = 2 case based on Gray codes. –Building block for larger parameters?

Open Questions Lots and lots of open questions. –Complexity of finding optimal designs for given parameters. –Asymptotically good codes for larger parameters. –Lower bounds. –Reasonable models for real systems. –Small “families” of codes good over ranges of different user behaviors. –Multi-objective: tradeoffs between average/worst- case performance. –Incorporating error-correction. –Extending to buffer codes, or other models. –And more.