consistent durable and safe memory management for byte
play

Consistent, Durable, and Safe Memory Management for Byte-Addressable - PowerPoint PPT Presentation

Apr 22, 2023 •105 likes •961 views

Consistent, Durable, and Safe Memory Management for Byte-Addressable Non-Volatile Main Memory Iulian Moraru*, David Andersen*, Michael Kaminsky, Niraj Tolia , Nathan Binkert, Partha Ranganathan *Carnegie Mellon

  1. ♢ ♢ Consistent, Durable, and Safe Memory Management for Byte-Addressable Non-Volatile Main Memory Iulian Moraru*, David Andersen*, Michael Kaminsky°, ♦ ‡ Niraj Tolia , Nathan Binkert, Partha Ranganathan ♦ ‡ *Carnegie Mellon University, °Intel Labs, Maginatics, Amiato, Google Sunday, November 3, 13

  2. New memory technologies 1 DRAM Flash x x 0 1 10 100 1000 10000 100000 Access time [ns] logscale 2 Sunday, November 3, 13

  3. New memory technologies NVRAM: phase change, spin torque, memristor • Memory that is both fast and non-volatile • Will help build fast, robust systems 1 DRAM Flash x x 0 1 10 100 1000 10000 100000 Access time [ns] logscale 2 Sunday, November 3, 13

  4. New memory technologies NVRAM: phase change, spin torque, memristor • Memory that is both fast and non-volatile • Will help build fast, robust systems NVRAM much faster than HDD and Flash 1 DRAM Flash x x 0 1 10 100 1000 10000 100000 Access time [ns] logscale 2 Sunday, November 3, 13

  5. New memory technologies NVRAM: phase change, spin torque, memristor • Memory that is both fast and non-volatile • Will help build fast, robust systems NVRAM much faster than HDD and Flash 1 DRAM Flash x x 0 1 10 100 1000 10000 100000 Access time [ns] logscale 2 Sunday, November 3, 13

  6. New memory technologies NVRAM: phase change, spin torque, memristor • Memory that is both fast and non-volatile • Will help build fast, robust systems NVRAM much faster than HDD and Flash 1 DRAM Flash x x x 0 1 10 100 1000 10000 100000 Access time PCM [ns] logscale 2 Sunday, November 3, 13

  7. New memory technologies NVRAM: phase change, spin torque, memristor • Memory that is both fast and non-volatile • Will help build fast, robust systems NVRAM much faster than HDD and Flash 1 DRAM Flash x x x 0 1 10 100 1000 10000 100000 Access time PCM [ns] logscale • Placed on the memory bus (alongside DRAM) 2 Sunday, November 3, 13

  8. New memory technologies NVRAM: phase change, spin torque, memristor • Memory that is both fast and non-volatile • Will help build fast, robust systems NVRAM much faster than HDD and Flash 1 DRAM Flash x x x 0 1 10 100 1000 10000 100000 Access time PCM [ns] logscale • Placed on the memory bus (alongside DRAM) • Needs new ways of handling persistent state 2 Sunday, November 3, 13

  9. NV main memory - assumptions Applications OS DRAM + NVRAM • Mapped into process address space • Accessed through CPU loads/stores • Persistent namespace 3 Sunday, November 3, 13

  10. NV main memory - assumptions Applications Persistence OS Library DRAM + NVRAM • Mapped into process address space • Accessed through CPU loads/stores • Persistent namespace 3 Sunday, November 3, 13

  11. NVRAM - challenge High performance and Robustness (no data loss) ➡ Give developers the right tools to develop data structures for NVRAM 4 Sunday, November 3, 13

  12. Sources of data loss 5 Sunday, November 3, 13

  13. Sources of data loss Wear-out 5 Sunday, November 3, 13

  14. Sources of data loss Wear-out Erroneous writes 5 Sunday, November 3, 13

  15. Sources of data loss Wear-out Erroneous writes CPU caches 5 Sunday, November 3, 13

  16. Sources of data loss Wear-out Erroneous writes CPU caches ... 5 Sunday, November 3, 13

  17. Sources of data loss Wear-out Memory allocator Erroneous writes CPU caches ... 5 Sunday, November 3, 13

  18. Sources of data loss Wear-out Memory allocator Virtual memory protection Erroneous writes CPU caches ... 5 Sunday, November 3, 13

  19. Sources of data loss Wear-out Memory allocator Virtual memory protection Erroneous writes CPU caches Cache line counters ... 5 Sunday, November 3, 13

  20. Malloc for NVRAM - goals 6 Sunday, November 3, 13

  21. Malloc for NVRAM - goals • Avoid frequently re-allocating the same block • HW wear leveling may slow down applications that write often to one location 6 Sunday, November 3, 13

  22. Malloc for NVRAM - goals • Avoid frequently re-allocating the same block • HW wear leveling may slow down applications that write often to one location • Minimize # of metadata updates in NVRAM • Wear-out & speed 6 Sunday, November 3, 13

  23. Malloc for NVRAM - goals • Avoid frequently re-allocating the same block • HW wear leveling may slow down applications that write often to one location • Minimize # of metadata updates in NVRAM • Wear-out & speed • Make metadata robust to accidental corruption • Avoids extensive loss of (persistent) data 6 Sunday, November 3, 13

  24. Problems with allocators for DRAM • Reuse recently freed blocks • More writes than necessary • Metadata is embedded in allocated / free blocks • Frequent writes to one location 1 1 1 7 Sunday, November 3, 13

  25. Problems with allocators for DRAM • Reuse recently freed blocks • More writes than necessary • Metadata is embedded in allocated / free blocks • Frequent writes to one location 3 1 1 1 7 Sunday, November 3, 13

  26. Problems with allocators for DRAM • Reuse recently freed blocks • More writes than necessary • Metadata is embedded in allocated / free blocks • Frequent writes to one location size 3 1 1 1 7 Sunday, November 3, 13

  27. Problems with allocators for DRAM • Reuse recently freed blocks • More writes than necessary • Metadata is embedded in allocated / free blocks • Frequent writes to one location size 3 1 2 1 1 1 7 Sunday, November 3, 13

  28. Problems with allocators for DRAM • Reuse recently freed blocks • More writes than necessary • Metadata is embedded in allocated / free blocks • Frequent writes to one location size 3 1 2 1 1 1 1 1 7 Sunday, November 3, 13

  29. Problems with allocators for DRAM • Reuse recently freed blocks • More writes than necessary • Metadata is embedded in allocated / free blocks • Frequent writes to one location size 3 1 2 1 1 1 1 1 1 7 Sunday, November 3, 13

  30. Problems with allocators for DRAM • Reuse recently freed blocks • More writes than necessary • Metadata is embedded in allocated / free blocks • Frequent writes to one location size 3 1 2 1 1 1 1 1 1 7 Sunday, November 3, 13

  31. Problems with allocators for DRAM • Reuse recently freed blocks • More writes than necessary • Metadata is embedded in allocated / free blocks • Frequent writes to one location size 1 3 1 1 2 1 1 1 7 Sunday, November 3, 13

  32. Problems with allocators for DRAM • Reuse recently freed blocks • More writes than necessary • Metadata is embedded in allocated / free blocks • Frequent writes to one location size 1 1 3 2 2 3 1 1 1 7 Sunday, November 3, 13

  33. Problems with allocators for DRAM • Reuse recently freed blocks • More writes than necessary • Metadata is embedded in allocated / free blocks • Frequent writes to one location size 1 1 1 3 2 2 2 3 3 1 1 1 7 Sunday, November 3, 13

  34. Problems with allocators for DRAM • Reuse recently freed blocks • More writes than necessary • Metadata is embedded in allocated / free blocks • Frequent writes to one location size 1 1 1 3 2 2 2 3 3 • Metadata corruption can cause data loss 1 1 1 7 Sunday, November 3, 13

  35. Problems with allocators for DRAM • Reuse recently freed blocks • More writes than necessary • Metadata is embedded in allocated / free blocks • Frequent writes to one location size 1 1 1 3 2 2 2 3 3 • Metadata corruption can cause data loss 1 1 1 7 Sunday, November 3, 13

  36. Problems with allocators for DRAM • Reuse recently freed blocks • More writes than necessary • Metadata is embedded in allocated / free blocks • Frequent writes to one location size 1 1 1 3 2 2 2 3 3 • Metadata corruption can cause data loss 1 3 1 1 7 Sunday, November 3, 13

  37. Problems with allocators for DRAM • Reuse recently freed blocks • More writes than necessary • Metadata is embedded in allocated / free blocks • Frequent writes to one location size 1 1 1 3 2 2 2 3 3 • Metadata corruption can cause data loss 1 3 7 Sunday, November 3, 13

  38. Problems with allocators for DRAM • Reuse recently freed blocks • More writes than necessary • Metadata is embedded in allocated / free blocks • Frequent writes to one location size 1 1 1 3 2 2 2 3 3 • Metadata corruption can cause data loss X X 1 3 7 Sunday, November 3, 13

  39. NVMalloc DRAM NVRAM 8 Sunday, November 3, 13

  40. NVMalloc DRAM NVRAM 8 Sunday, November 3, 13

  41. NVMalloc free / in-use bitmap 000000 ... DRAM NVRAM 8 Sunday, November 3, 13

  42. NVMalloc free / in-use bitmap 000000 ... 010001 ... DRAM NVRAM 8 Sunday, November 3, 13

  43. NVMalloc free lists ... free / in-use bitmap size 1: 000000 ... 010001 ... ... size 2: ... size 3: DRAM NVRAM 8 Sunday, November 3, 13

  44. NVMalloc free lists ... free / in-use bitmap size 1: 000000 ... 010001 ... ... size 2: ... size 3: DRAM NVRAM 8 Sunday, November 3, 13

  45. NVMalloc free lists ... free / in-use bitmap size 1: 000000 ... 010001 ... ... size 2: ... size 3: DRAM NVRAM 8 Sunday, November 3, 13

Recommend

Consistent and Durable Data Structures for Non-Volatile Byte-Addressable Memory Shivaram

Consistent and Durable Data Structures for Non-Volatile Byte-Addressable Memory Shivaram

Consistent and Durable Data Structures for Non-Volatile Byte-Addressable Memory Shivaram Venkataraman* , Niraj Tolia , Parthasarathy Ranganathan* and Roy H. Campbell *HP Labs, Palo Alto, Maginatics, and University of Illinois,

422 views • 27 slides
Memory Management Memory Management Memory a linear array of bits, bytes, words, pages ... -

Memory Management Memory Management Memory a linear array of bits, bytes, words, pages ... -

CS510 Operating System Foundations Jonathan Walpole Memory Management Memory Management Memory a linear array of bits, bytes, words, pages ... - Each byte is named by a unique memory address - Holds instructions and data for OS and user

761 views • 56 slides
TLBs 1 memory HW random memory image page tables with 1-byte page entries answer: 2-byte

TLBs 1 memory HW random memory image page tables with 1-byte page entries answer: 2-byte

TLBs 1 memory HW random memory image page tables with 1-byte page entries answer: 2-byte values read (or replaced) or fault 3 attempts per set of problems submitting only right and blank answers doesnt count as attempt keep

1.2k views • 95 slides
CS333 Intro to Operating Systems Jonathan Walpole Memory Management Memory Management Memory

CS333 Intro to Operating Systems Jonathan Walpole Memory Management Memory Management Memory

CS333 Intro to Operating Systems Jonathan Walpole Memory Management Memory Management Memory a linear array of bytes - Holds O.S. and programs (processes) - Each cell (byte) is named by a unique memory address Recall, processes are

576 views • 55 slides
multi-byte values in memory Store across contiguous byte locations. 64-bit Words Bytes Address

multi-byte values in memory Store across contiguous byte locations. 64-bit Words Bytes Address

multi-byte values in memory Store across contiguous byte locations. 64-bit Words Bytes Address 0x1F 0x1E Programming with Memory Alignment (Why?) 0x1D 0x1C 0x1B via C, pointers, and arrays 0x1A 0x19 0x18 0x17 0x16 0x15 0x14 0x13

487 views • 5 slides
Data, memory, pointer Pointers and arrays 1-1 Data, memory memory address: every byte

Data, memory, pointer Pointers and arrays 1-1 Data, memory memory address: every byte

Data, memory, pointer Pointers and arrays 1-1 Data, memory memory address: every byte is identified by a numeric address in the memory. once a variable is defined , one cannot predict its memory address, entirely determined by

976 views • 70 slides
Safe, consistent and effective management of Atrial Fibrillation in the Emergency Department

Safe, consistent and effective management of Atrial Fibrillation in the Emergency Department

The day my heart skipped a beat Safe, consistent and effective management of Atrial Fibrillation in the Emergency Department Objectives The problems The team The interventions Results Learning The Patient The

542 views • 19 slides
Better I/O Through Byte-Addressable, Persistent Memory Jeremy Condit , Ed Nightingale, Chris

Better I/O Through Byte-Addressable, Persistent Memory Jeremy Condit , Ed Nightingale, Chris

Better I/O Through Byte-Addressable, Persistent Memory Jeremy Condit , Ed Nightingale, Chris Frost, Engin Ipek, Ben Lee, Doug Burger, Derrick Coetzee A New World of Storage DRAM + Fast + Byte-addressable - Volatile Disk / Flash +

907 views • 71 slides
Durable Transactional Memory Can Scale With TimeStone * R. Madhava Krishnan , Jaeho Kim * , Ajit

Durable Transactional Memory Can Scale With TimeStone * R. Madhava Krishnan , Jaeho Kim * , Ajit

Durable Transactional Memory Can Scale With TimeStone * R. Madhava Krishnan , Jaeho Kim * , Ajit Mathew, Xinwei Fu, Anthony Demeri, Changwoo Min, Sudarsun Kannan + + Executive Summary TimeStone is a highly scalable Durable Transaction

705 views • 56 slides
Memory Management Memory Management 5A. Memory Management and Address Spaces 1. allocate/assign

Memory Management Memory Management 5A. Memory Management and Address Spaces 1. allocate/assign

4/14/2018 Memory Management Memory Management 5A. Memory Management and Address Spaces 1. allocate/assign physical memory to processes 5B. Memory Allocation explicit requests: malloc (sbrk) implicit: program loading, stack extension

555 views • 9 slides
Memory Management Memory Management 5A. Memory Management and Address Spaces 1. allocate/assign

Memory Management Memory Management 5A. Memory Management and Address Spaces 1. allocate/assign

4/12/2017 Memory Management Memory Management 5A. Memory Management and Address Spaces 1. allocate/assign physical memory to processes 5B. Simple Memory Allocation explicit requests: malloc (sbrk) implicit: program loading, stack

205 views • 8 slides
Motivation Organize main memory to Facilitate byte-addressability EE 457 Unit 7b while

Motivation Organize main memory to Facilitate byte-addressability EE 457 Unit 7b while

7b.1 7b.2 Motivation Organize main memory to Facilitate byte-addressability EE 457 Unit 7b while maintaining Efficient fetching of the words in a cache block __________________________ helps us achieve this Main Memory

222 views • 8 slides
The Memory Abstraction Association of <name, value> pairs typically named as byte

The Memory Abstraction Association of <name, value> pairs typically named as byte

The Memory Abstraction Association of <name, value> pairs typically named as byte addresses often values aligned on multiples of size Sequence of Reads and Writes Write binds a value to an address Read of addr returns

1.57k views • 121 slides
Lecture 5: Memory Management 1 / 54 Memory Management Administrivia Assignment 1 is due on

Lecture 5: Memory Management 1 / 54 Memory Management Administrivia Assignment 1 is due on

Memory Management Lecture 5: Memory Management 1 / 54 Memory Management Administrivia Assignment 1 is due on September 7th @ 11:59pm 2 / 54 Memory Management Poll https: // www.strawpoll.me / 20863490 3 / 54 Memory Management Recap

952 views • 54 slides
DHTM: Durable Hardware Transactional Memory Arpit Joshi , Vijay Nagarajan, Marcelo Cintra, Stratis

DHTM: Durable Hardware Transactional Memory Arpit Joshi , Vijay Nagarajan, Marcelo Cintra, Stratis

DHTM: Durable Hardware Transactional Memory Arpit Joshi , Vijay Nagarajan, Marcelo Cintra, Stratis Viglas ISCA 2018 Persistent Memory is here 2 Persistent Memory is here 2 Persistent Memory Systems L1 L1 LLC Persistent Memory

1.1k views • 73 slides
Future Memory Technologies Seminar WS2012/13 Benjamin Klenk 2013/02/08 Supervisor: Prof. Dr.

Future Memory Technologies Seminar WS2012/13 Benjamin Klenk 2013/02/08 Supervisor: Prof. Dr.

Future Memory Technologies Seminar WS2012/13 Benjamin Klenk 2013/02/08 Supervisor: Prof. Dr. Holger Frning Department of Computer Engineering University of Heidelberg 1 Amdahls rule of thumb 1 byte of memory and 1 byte per second of I/O

612 views • 43 slides
1 3. Simple Memory System Cache Address Translation Example #1 16 lines, 4-byte block size

1 3. Simple Memory System Cache Address Translation Example #1 16 lines, 4-byte block size

Today Simple memory system example Case study: Core i7/Linux memory system Virtual Memory: Systems Memory mapping CSci 2021: Machine Architecture and Organization April 20th-22nd, 2020 Your instructor: Stephen McCamant Based on

587 views • 5 slides
28.05.04 09:50 Memory Management The computer memory is a limited resource so the Memory

28.05.04 09:50 Memory Management The computer memory is a limited resource so the Memory

28.05.04 09:50 Memory Management The computer memory is a limited resource so the Memory Management memory use of programs has to be managed in some way. Memory Management The memory management is usually performed by a runtime system

782 views • 13 slides
FlashTier: A Lightweight, Consistent and Durable Storage Cache 1 Outline Introduction

FlashTier: A Lightweight, Consistent and Durable Storage Cache 1 Outline Introduction

467 views • 33 slides
Durable Business Drives Cash Flow and Supports Dividend Growth November 15-16, 2016 2 Safe

Durable Business Drives Cash Flow and Supports Dividend Growth November 15-16, 2016 2 Safe

Durable Business Drives Cash Flow and Supports Dividend Growth November 15-16, 2016 2 Safe Harbor Language and Reconciliation of Non-GAAP Measures Safe Harbor Statement Under the Private Securities Litigation Reform Act of 1995: Certain

916 views • 54 slides
Durable Business Drives Cash Flow and Supports Dividend Growth September 9, 2016 2 Safe

Durable Business Drives Cash Flow and Supports Dividend Growth September 9, 2016 2 Safe

Durable Business Drives Cash Flow and Supports Dividend Growth September 9, 2016 2 Safe Harbor Language and Reconciliation of Non-GAAP Measures Safe Harbor Statement Under the Private Securities Litigation Reform Act of 1995: Certain

919 views • 55 slides
Durable Business Drives Cash Flow and Supports Dividend Growth September 20-21, 2016 2 Safe

Durable Business Drives Cash Flow and Supports Dividend Growth September 20-21, 2016 2 Safe

Durable Business Drives Cash Flow and Supports Dividend Growth September 20-21, 2016 2 Safe Harbor Language and Reconciliation of Non-GAAP Measures Safe Harbor Statement Under the Private Securities Litigation Reform Act of 1995: Certain

725 views • 55 slides
Durable Business Drives Cash Flow and Dividend Growth June 2019 Safe Harbor Language and

Durable Business Drives Cash Flow and Dividend Growth June 2019 Safe Harbor Language and

Durable Business Drives Cash Flow and Dividend Growth June 2019 Safe Harbor Language and Reconciliation of 2 Non-GAAP Measures Forward Looking Statements Safe Harbor Statement Under the Private Securities Litigation Reform Act of 1995:

532 views • 37 slides
Durable Business Drives Cash Flow and Supports Dividend Growth August 18, 2016 2 Safe Harbor

Durable Business Drives Cash Flow and Supports Dividend Growth August 18, 2016 2 Safe Harbor

Durable Business Drives Cash Flow and Supports Dividend Growth August 18, 2016 2 Safe Harbor Language and Reconciliation of Non-GAAP Measures Safe Harbor Statement Under the Private Securities Litigation Reform Act of 1995: Certain statements

742 views • 57 slides

More recommend