raft and other stories
play

Raft and Other Stories Consensus Trilogy: Part III Rough Timeline - PowerPoint PPT Presentation

Jun 15, 2023 •207 likes •1.73k views

Raft and Other Stories Consensus Trilogy: Part III Rough Timeline for Today Talk about logistics and lab 0 Raft: You should all know this better than I do. Quiz and midterm course evaluation. Lab 0 and 2 Some Observations Nearly

  1. A Possible Solution set("s1", 42) set("s1", 42) set("s1", 25) set("s1", 25) set("s1", 1729) set("s1", 1729) KV-Store KV-Store Client Client Client set("s1", 25) set("s1", 42) set("s1", 1729)

  2. A Possible Solution set("s1", 42) set("s1", 42) KV-Store KV-Store set("s1", 25) set("s1", 25) set("s1", 1729) set("s1", 1729) Client Client Client set("s1", 25) set("s1", 42) set("s1", 1729)

  3. Raft Observation • Paxos originally explained as a single decree leaderless protocol.

  4. Raft Observation • Paxos originally explained as a single decree leaderless protocol. • Leader added for efficiency, e.g., in MultiPaxos from last lecture.

  5. Raft Observation • Paxos originally explained as a single decree leaderless protocol. • Leader added for efficiency, e.g., in MultiPaxos from last lecture. • Reduces conflicts. Reduces number of messages in the common case.

  6. Raft Observation • Paxos originally explained as a single decree leaderless protocol. • Leader added for efficiency, e.g., in MultiPaxos from last lecture. • Reduces conflicts. Reduces number of messages in the common case. • But leader election is independent of actual protocol.

  7. Raft Observation • Paxos originally explained as a single decree leaderless protocol. • Leader added for efficiency, e.g., in MultiPaxos from last lecture. • Reduces conflicts. Reduces number of messages in the common case. • But leader election is independent of actual protocol. • Raft combines these two processes in one protocol.

  8. Raft Observation • Paxos originally explained as a single decree leaderless protocol. • Leader added for efficiency, e.g., in MultiPaxos from last lecture. • Reduces conflicts. Reduces number of messages in the common case. • But leader election is independent of actual protocol. • Raft combines these two processes in one protocol. • Leader election responsible for ensuring a good leader.

  9. Raft Observation • Paxos originally explained as a single decree leaderless protocol. • Leader added for efficiency, e.g., in MultiPaxos from last lecture. • Reduces conflicts. Reduces number of messages in the common case. • But leader election is independent of actual protocol. • Raft combines these two processes in one protocol. • Leader election responsible for ensuring a good leader. • Consensus is just leader deciding on ordering and replicating decision.

  10. Raft Log • Primitive: Maintain a consistent log across participants

  11. Raft Log Index Term Command • Primitive: Maintain a consistent log across participants

  12. Raft Log Index Entries with the same term from the same leader. Term Command • Primitive: Maintain a consistent log across participants

  13. Raft Log Index Entries with the same term from the same leader. Term Command 0 0 set(x, 5) • Primitive: Maintain a consistent log across participants

  14. Raft Log Index Entries with the same term from the same leader. Term Command 0 1 0 0 set(x, 5) set(x, 6) • Primitive: Maintain a consistent log across participants

  15. Raft Log Index Entries with the same term from the same leader. Term Command 0 1 2 0 0 1 set(x, 5) set(x, 6) get(x) • Primitive: Maintain a consistent log across participants

  16. Raft Log Index Entries with the same term from the same leader. Term Command 0 1 2 3 0 0 1 0 set(x, 5) set(x, 6) get(x) get(x) • Primitive: Maintain a consistent log across participants

  17. Raft Log Index Entries with the same term from the same leader. Term Command 0 1 2 3 Terms should not go back 0 0 1 0 set(x, 5) set(x, 6) get(x) get(x) • Primitive: Maintain a consistent log across participants

  18. Raft Log Index Term Command 0 1 2 0 0 1 set(x, 5) set(x, 6) get(x) 0 1 2 0 0 1 set(x, 5) set(x, 6) get(x) 0 1 2 0 0 1 set(x, 5) set(x, 6) set(z, 2)

  19. Raft Log Index Term Command 0 1 2 0 0 1 set(x, 5) set(x, 6) get(x) 0 1 2 Two entries with the same index and term are identical. 0 0 1 set(x, 5) set(x, 6) get(x) 0 1 2 0 0 1 set(x, 5) set(x, 6) set(z, 2)

  20. Raft Log Index Term Command 0 1 2 0 0 1 set(x, 5) set(x, 6) get(x) 0 1 2 0 0 1 set(x, 5) set(x, 6) get(x) 0 1 2 0 0 0 set(x, 5) set(x, 6) set(z, 8)

  21. Raft Log Index Term Command 0 1 2 3 0 0 1 1 set(x, 5) set(x, 6) get(x) set(y, 6) 0 1 2 0 0 1 set(x, 5) set(x, 6) get(x) 0 1 2 0 0 0 set(x, 5) set(x, 6) set(z, 8)

  22. Raft Log Index Term Command 0 1 2 3 0 0 1 1 set(x, 5) set(x, 6) get(x) set(y, 6) 0 1 2 3 0 0 1 1 set(x, 5) set(x, 6) get(x) set(y, 6) 0 1 2 0 0 0 set(x, 5) set(x, 6) set(z, 8)

  23. Raft Log Index Term Command 0 1 2 3 0 0 1 1 set(x, 5) set(x, 6) get(x) set(y, 6) 0 1 2 3 0 0 1 1 set(x, 5) set(x, 6) get(x) set(y, 6) 0 1 2 3 0 0 0 1 set(x, 5) set(x, 6) set(z, 8) set(y, 6)

  24. Raft Log Index Term Command 0 1 2 3 0 0 1 1 set(x, 5) set(x, 6) get(x) set(y, 6) 0 1 2 3 If two logs agree on an entry, they agree on the prefix 0 0 1 1 set(x, 5) set(x, 6) get(x) set(y, 6) 0 1 2 3 0 0 0 1 set(x, 5) set(x, 6) set(z, 8) set(y, 6)

  25. Raft Log Index Term Command 0 1 2 3 0 0 1 1 set(x, 5) set(x, 6) get(x) set(y, 6) 0 1 2 3 If two logs agree on an entry, they agree on the prefix 0 0 1 1 set(x, 5) set(x, 6) get(x) set(y, 6) 0 1 2 0 0 0 set(x, 5) set(x, 6) set(z, 8)

  26. Raft Log Index Term Command 0 1 2 3 0 0 1 1 set(x, 5) set(x, 6) get(x) set(y, 6) 0 1 2 3 If two logs agree on an entry, they agree on the prefix 0 0 1 1 set(x, 5) set(x, 6) get(x) set(y, 6) 0 1 0 0 set(x, 5) set(x, 6)

  27. Raft Log Index Term Command 0 1 2 3 0 0 1 1 set(x, 5) set(x, 6) get(x) set(y, 6) 0 1 2 3 If two logs agree on an entry, they agree on the prefix 0 0 1 1 set(x, 5) set(x, 6) get(x) set(y, 6) 0 1 2 0 0 1 set(x, 5) set(x, 6) get(x)

  28. Raft Log Index Term Command 0 1 2 3 0 0 1 1 set(x, 5) set(x, 6) get(x) set(y, 6) 0 1 2 3 If two logs agree on an entry, they agree on the prefix 0 0 1 1 set(x, 5) set(x, 6) get(x) set(y, 6) 0 1 2 3 0 0 1 1 set(x, 5) set(x, 6) get(x) set(y, 6)

  29. Raft Timeline Term

  30. Raft Timeline n o i t c e l E r e d a e L Term

  31. L e a d Term e r R E e l i e g c n t i o n Raft Timeline

  32. L e a d Term e r R E e l i e g c n t i o n L e a d e r E l e c t i o n Raft Timeline

  33. L e a d Term e r R E e l i e g c n t i o n L e a d e r E l e c t i o n Raft Timeline

  34. L e a d Term e r R E e l i e g c n t i o n L e a d e r E l e c t i o n Raft Timeline

  35. Raft within a Term 0 0 set(x, 5) 0 0 set(x, 5) set(x, 6)

  36. Raft within a Term 0 0 set(x, 5) AppendEntries(term=0, prevIdx =0, prevTerm=0, [(1,0, set(x,6)]) 0 1 0 0 set(x, 5) set(x, 6) AppendEntries(term=0, prevIdx =0, prevTerm=0, [(1,0, set(x,6)])

  37. Raft within a Term 0 0 set(x, 5) AppendEntries(term=0, prevIdx =0, prevTerm=0, [(1,0, set(x,6)]) 0 1 How should the followers respond? 0 0 set(x, 5) set(x, 6) AppendEntries(term=0, prevIdx =0, prevTerm=0, [(1,0, set(x,6)])

  38. Raft within a Term 0 1 0 0 set(x, 5) set(x, 6) (0, Success) 0 1 0 0 set(x, 5) set(x, 6) (0, Fail)

  39. Raft within a Term 0 1 0 0 set(x, 5) set(x, 6) (0, Success) 0 1 What can the leader infer from this interaction? 0 0 set(x, 5) set(x, 6) (0, Fail)

  40. Raft within a Term 0 1 0 0 set(x, 5) set(x, 6) 0 1 0 0 set(x, 5) set(x, 6) AppendEntries(term=-1, prevIdx=-1, prevTerm=-1, [(0,0, set(x, 5),(1,0, set(x,6)])

  41. Raft within a Term 0 1 0 0 set(x, 5) set(x, 6) 0 1 0 0 set(x, 5) set(x, 6) 0 1 0 0 set(x, 5) set(x, 6)

  42. Raft within a Term 0 1 ... 0 0 set(x, 5) set(x, 6) 0 1 22 23 24 ... 0 0 0 0 0 set(x, 5) set(x, 6) ... ... ... 0 1 ... 0 0 set(x, 5) set(x, 6)

  43. Raft within a Term 0 1 ... 0 0 set(x, 5) set(x, 6) 0 1 22 23 24 ... 0 0 0 0 0 set(x, 5) set(x, 6) ... ... ... 0 1 ... 0 0 set(x, 5) set(x, 6)

  44. Raft within a Term 0 1 ... 0 0 set(x, 5) set(x, 6) 0 1 22 23 24 ... 0 0 0 0 0 set(x, 5) set(x, 6) ... ... ... 0 1 22 ... 0 0 1 set(x, 5) set(x, 6) ...

  45. Raft within a Term 0 1 22 ... 0 0 1 set(x, 5) set(x, 6) ... 0 1 22 23 24 ... 0 0 0 0 0 set(x, 5) set(x, 6) ... ... ... 0 1 22 ... 0 0 1 set(x, 5) set(x, 6) ...

  46. Raft within a Term 0 1 22 ... 0 0 1 set(x, 5) set(x, 6) ... 0 1 ... 0 0 set(x, 5) set(x, 6) 0 1 22 ... 1 0 0 set(x, 5) set(x, 6) ...

  47. Raft within a Term 0 1 22 ... 0 0 1 set(x, 5) set(x, 6) ... 0 1 22 ... 0 0 1 set(x, 5) set(x, 6) ... 0 1 22 ... 1 0 0 set(x, 5) set(x, 6) ...

  48. Raft within a Term 0 1 22 ... 0 0 1 set(x, 5) set(x, 6) ... 0 1 22 ... 0 0 1 set(x, 5) set(x, 6) ... Log entries can be deleted. When is it safe to execute commands? 0 1 22 ... 1 0 0 set(x, 5) set(x, 6) ...

  49. Need to ensure some log entries are never deleted.

  50. Safety Through Leader Election • Want to make sure some set of entries always remain in the log.

  51. Safety Through Leader Election • Want to make sure some set of entries always remain in the log. • Can only execute actions for such committed entries.

  52. Safety Through Leader Election • Want to make sure some set of entries always remain in the log. • Can only execute actions for such committed entries. • Need to ensure this without communicating with former leader(s).

  53. Safety Through Leader Election • Want to make sure some set of entries always remain in the log. • Can only execute actions for such committed entries. • Need to ensure this without communicating with former leader(s). • Leaders might die or disappear.

Recommend

L-102.00 1 Building Outline Map L-102.00 SCALE: 1" = 10' Date: 7/17/2017 Date:

L-102.00 1 Building Outline Map L-102.00 SCALE: 1" = 10' Date: 7/17/2017 Date:

92ND STREET 7 4 4 4 4 Stories Stories Stories Stories Stories 7 Stories 15 5 4 4 4 4 4 Stories Stories Stories Stories Stories Stories Stories 4 Stories 3 Historic Photo A 2 Historic Photo A L-102.00 L-102.00 SCALE:

477 views • 5 slides
Rosalind Dibley: Digital Stories and Me Sharing stories Information about how to tell stories is

Rosalind Dibley: Digital Stories and Me Sharing stories Information about how to tell stories is

Rosalind Dibley: Digital Stories and Me Sharing stories Information about how to tell stories is everywhere!!!!!!!!!! ! www.storycenter.org Stories that come from official sources, such as ne news me ws media dia co corpo poratio

489 views • 22 slides
Consensus II Replicated State Machines, RAFT CS 240: Computing Systems and Concurrency Lecture

Consensus II Replicated State Machines, RAFT CS 240: Computing Systems and Concurrency Lecture

Consensus II Replicated State Machines, RAFT CS 240: Computing Systems and Concurrency Lecture 10 Marco Canini Credits: Michael Freedman and Kyle Jamieson developed much of the original material. RAFT slides heavily based on those from Diego

437 views • 41 slides
West 108th Street Development West Side Federation for Senior & Supportive Housing Dattner

West 108th Street Development West Side Federation for Senior & Supportive Housing Dattner

WEST 109TH STREET COLUMBUS AVE. AMSTERDAM AVE. ANABIL AVILES PLAYGROUND 6 Stories PLAYGROUND 11 Stories 11 Stories TOILETS 9 Stories 7 Stories 7 Stories WEST 108TH STREET BUILDING 1 BUILDING 2 BOOKER T. WASHINGTON M.S. 54

310 views • 5 slides
Designing for Understandability: the Raft Consensus Algorithm Diego Ongaro John Ousterhout

Designing for Understandability: the Raft Consensus Algorithm Diego Ongaro John Ousterhout

Designing for Understandability: the Raft Consensus Algorithm Diego Ongaro John Ousterhout Stanford University Algorithms Should Be Designed For ... Correctness? Efficiency? Conciseness? Understandability! August 29, 2016 The Raft

1.55k views • 29 slides
THE RAFT Resilience Adaptation Feasibility Tool PRESENTATION TO NORTHAMPTON COUNTY PLANNING

THE RAFT Resilience Adaptation Feasibility Tool PRESENTATION TO NORTHAMPTON COUNTY PLANNING

THE RAFT Resilience Adaptation Feasibility Tool PRESENTATION TO NORTHAMPTON COUNTY PLANNING COMMISSION JANUARY 16, 2019 THE RAFT GOAL To help Virginias coastal localities improve resilience to flooding and other coastal storm hazards

452 views • 34 slides
TELLING STORIES TECHNOLOGY & MEDIA DREW DAVIDSON A LITTLE BIT ABOUT ME STORIES

TELLING STORIES TECHNOLOGY & MEDIA DREW DAVIDSON A LITTLE BIT ABOUT ME STORIES

TELLING STORIES TECHNOLOGY & MEDIA DREW DAVIDSON A LITTLE BIT ABOUT ME STORIES Fairytales, Folktales, Myths, Legends, Fables Narratives, Fictions, Tales, Dramas Once upon a time Everyday But one day Because of this

986 views • 76 slides
P RESENTATION TO THE I NTERNATIONAL S YMPOSIUM O N THE D RAFT A NTI -M ONOPOLY L AW OF THE P EOPLE

P RESENTATION TO THE I NTERNATIONAL S YMPOSIUM O N THE D RAFT A NTI -M ONOPOLY L AW OF THE P EOPLE

P RESENTATION TO THE I NTERNATIONAL S YMPOSIUM O N THE D RAFT A NTI -M ONOPOLY L AW OF THE P EOPLE S R EPUBLIC OF C HINA Legislative Affairs Office of the State Council May 23 and 24, 2005 Beijing William Blumenthal * General Counsel

414 views • 16 slides
Dra raft ft Com omms Data ta Bil ill UK UKNOF OF 9/ 9/10 10/2 /201 012 Disc scla

Dra raft ft Com omms Data ta Bil ill UK UKNOF OF 9/ 9/10 10/2 /201 012 Disc scla

Dra raft ft Com omms Data ta Bil ill UK UKNOF OF 9/ 9/10 10/2 /201 012 Disc scla laime imer: r: this is my distillatio illation n of 448 pages s of detailed consultation responses. Ive tried to produce a fair and bala

742 views • 33 slides
Bitcoin & RAFT Distributed Systems Nikita Borisov Topics for Today Finish Bitcoin

Bitcoin & RAFT Distributed Systems Nikita Borisov Topics for Today Finish Bitcoin

Bitcoin & RAFT Distributed Systems Nikita Borisov Topics for Today Finish Bitcoin Broadcast mechanism Overview of MP2 Raft consensus Bitcoin broadcast Need to broadcast: Transactions to all nodes, so they can be

453 views • 33 slides
LOD Stories { Learning About Art by Building Multimedia Stories Hao Zhang, Jianliang Chen,

LOD Stories { Learning About Art by Building Multimedia Stories Hao Zhang, Jianliang Chen,

LOD Stories { Learning About Art by Building Multimedia Stories Hao Zhang, Jianliang Chen, Yuting Liu, Dipanwita Maulik, Linda Xu Dr. Craig A. Knoblock, Dr. Pedro Szekely University of Southern California, United States MielVander Sande Ghent

743 views • 52 slides
Data and potatoes (some drafts of stories) NomenclatureS Stories that we could have told

Data and potatoes (some drafts of stories) NomenclatureS Stories that we could have told

Data and potatoes (some drafts of stories) NomenclatureS Stories that we could have told Some graphes and stories Nomenclatures 1. Central Product Classification 2. Dfinition and classification of commodities

510 views • 18 slides
User Stories UC Santa Barbara Similar to Use Cases but not the same User stories are

User Stories UC Santa Barbara Similar to Use Cases but not the same User stories are

From Prof. Krintz, Winter 2018 User Stories UC Santa Barbara Similar to Use Cases but not the same User stories are centered on the result and the benefit of the thing youre describing, whereas use cases are more granular, and describe

56 views • 5 slides
Benchmarking Methodology for IPv6 Transition Technologies

Benchmarking Methodology for IPv6 Transition Technologies

iplab Benchmarking Methodology for IPv6 Transition Technologies draft-ietf-bmwg-ipv6-tran-tech-benchmarking-01 Marius Georgescu Nara Institute of Science and Technology Internet Engineering Laboratory 7 Apr. 2016 D RAFT M OTIVATION D RAFT

340 views • 13 slides
stories in between a presentatio n drew davidso n stories and mediums across and betwee n w ritten

stories in between a presentatio n drew davidso n stories and mediums across and betwee n w ritten

stories in between a presentatio n drew davidso n stories and mediums across and betwee n w ritten word sequential ar t hypermedia objects of study mys t hypermedia, novels, comic sandma n ultima onlin e mitternachtspiel storytelling

336 views • 14 slides
The Power and Paradox of Patient Stories Patient stories can: Provide a human face to your

The Power and Paradox of Patient Stories Patient stories can: Provide a human face to your

The Power and Paradox of Patient Stories Patient stories can: Provide a human face to your Draw unwanted judgements narrative Increase stigma Attract the interest of journalists and the public Create an unrealistic

302 views • 16 slides
How choosing the Raft consensus algorithm saved us 3 months of development time What do I do

How choosing the Raft consensus algorithm saved us 3 months of development time What do I do

How choosing the Raft consensus algorithm saved us 3 months of development time What do I do with unused space on my servers? Lets build an S3 cluster! Requirements: Fully S3 compatible Easy to maintain Fault tolerant I found a

615 views • 38 slides
Welc lcom ome! The D The Dra raft G Gra rand T Tru runk Ma Master Plan an Open Hou Op

Welc lcom ome! The D The Dra raft G Gra rand T Tru runk Ma Master Plan an Open Hou Op

Welc lcom ome! The D The Dra raft G Gra rand T Tru runk Ma Master Plan an Open Hou Op ouse No. No. 2 2 NOV OVEMB MBER 7 7, 2017 17 Present ntat atio ion A Agend nda Overview presentation Project recap Where

510 views • 35 slides
PAL Stories of Impact Sarah Bailey Deputy Director of SEWS We cant help everyone, but

PAL Stories of Impact Sarah Bailey Deputy Director of SEWS We cant help everyone, but

PAL Stories of Impact Sarah Bailey Deputy Director of SEWS We cant help everyone, but everyone can help someone. Summary The PAL story Peer Mentoring stories Peer Assisted Study stories Questions Chapter 1 The

478 views • 28 slides
Edmund Coleman-Fountain Janice McLaughlin Stories and Practices Relationship between practices

Edmund Coleman-Fountain Janice McLaughlin Stories and Practices Relationship between practices

Re-Shaping Young Disabled Bodies: Stories of medical and self-intervention Edmund Coleman-Fountain Janice McLaughlin Stories and Practices Relationship between practices and stories: Disabled young body reshaped by stories and practices.

457 views • 11 slides
D RAFT P REFERRED S CENARIO : O VERVIEW OF G ROWTH P ATTERN & I NVESTMENT S TRATEGY Image

D RAFT P REFERRED S CENARIO : O VERVIEW OF G ROWTH P ATTERN & I NVESTMENT S TRATEGY Image

D RAFT P REFERRED S CENARIO : O VERVIEW OF G ROWTH P ATTERN & I NVESTMENT S TRATEGY Image Source: https://www.flickr.com/photos/adamrschultz/8810617814 Ken Kirkey & Matt Maloney, MTC September 20, 2016 San Mateo County Planning

753 views • 29 slides
PERFORMANCE OF FLOATING PILED RAFT WITH VARYING LENGTHS IN SOFT COMPRESSIBLE SUBSOIL By : Ir.

PERFORMANCE OF FLOATING PILED RAFT WITH VARYING LENGTHS IN SOFT COMPRESSIBLE SUBSOIL By : Ir.

1 PERFORMANCE OF FLOATING PILED RAFT WITH VARYING LENGTHS IN SOFT COMPRESSIBLE SUBSOIL By : Ir. Dr. TAN Yean Chin G&P Professionals Sdn Bhd Email : yctan@gnpgroup.com.my 25 June 2018 2 Contents Introduction Problem Statements

1.28k views • 98 slides
Traceable Anonymous Certificate dra raft-ie ietf-pkix-tac-01.txt IE IETF-72 at t PKIX IX WG

Traceable Anonymous Certificate dra raft-ie ietf-pkix-tac-01.txt IE IETF-72 at t PKIX IX WG

Traceable Anonymous Certificate dra raft-ie ietf-pkix-tac-01.txt IE IETF-72 at t PKIX IX WG Park, SangHwan shpark@kisa.or.kr Stephen Kent kent@bbn.com Overview I-D defines a practical architecture and protocols for offering privacy in

275 views • 8 slides
By The Life Raft Group What is cancer? Cancer consists of over 100 diseases. Cancer begins

By The Life Raft Group What is cancer? Cancer consists of over 100 diseases. Cancer begins

By The Life Raft Group What is cancer? Cancer consists of over 100 diseases. Cancer begins when there is an abnormal growth of cells that cannot be controlled. Because different types of cancers behave differently, it is important to

783 views • 50 slides

More recommend