57
TRUST Spring Conference, April 2-3, 2008 Write Markers for Probabilistic Quorum Systems Michael Merideth, Carnegie Mellon University Michael Reiter, University of North Carolina

TRUST Spring Conference, April 2-3, 2008 Write Markers for Probabilistic Quorum Systems Michael Merideth, Carnegie Mellon University Michael Reiter, University

Embed Size (px)

Citation preview

Page 1: TRUST Spring Conference, April 2-3, 2008 Write Markers for Probabilistic Quorum Systems Michael Merideth, Carnegie Mellon University Michael Reiter, University

TRUST Spring Conference, April 2-3, 2008

Write Markers for Probabilistic Quorum Systems

Michael Merideth, Carnegie Mellon University

Michael Reiter, University of North Carolina

Page 2: TRUST Spring Conference, April 2-3, 2008 Write Markers for Probabilistic Quorum Systems Michael Merideth, Carnegie Mellon University Michael Reiter, University

4/3/08 Michael Merideth 2

Replication via Quorum Systems

Replicated data– Server becomes n replicas

Server

Clients

Page 3: TRUST Spring Conference, April 2-3, 2008 Write Markers for Probabilistic Quorum Systems Michael Merideth, Carnegie Mellon University Michael Reiter, University

4/3/08 Michael Merideth 3

Replicas Replicated data

– Server becomes n replicas Clients issue read and write

operations– Involve quorums (subsets)

of replicas

High availability– Yet, no writes lost,

forged, or corrupted Clients

Replication via Quorum Systems

Page 4: TRUST Spring Conference, April 2-3, 2008 Write Markers for Probabilistic Quorum Systems Michael Merideth, Carnegie Mellon University Michael Reiter, University

4/3/08 Michael Merideth 4

Types of Servers (in Examples)

bowling ball ice cream fish any value

non-faulty faulty

Page 5: TRUST Spring Conference, April 2-3, 2008 Write Markers for Probabilistic Quorum Systems Michael Merideth, Carnegie Mellon University Michael Reiter, University

4/3/08 Michael Merideth 5

Types of Clients (in Examples)

non-faulty faulty

Page 6: TRUST Spring Conference, April 2-3, 2008 Write Markers for Probabilistic Quorum Systems Michael Merideth, Carnegie Mellon University Michael Reiter, University

4/3/08 Michael Merideth 6

Write Operation

Client wants to write “ice cream” to system

Page 7: TRUST Spring Conference, April 2-3, 2008 Write Markers for Probabilistic Quorum Systems Michael Merideth, Carnegie Mellon University Michael Reiter, University

4/3/08 Michael Merideth 7

Write Operation

Client submits write to write quorum

Page 8: TRUST Spring Conference, April 2-3, 2008 Write Markers for Probabilistic Quorum Systems Michael Merideth, Carnegie Mellon University Michael Reiter, University

4/3/08 Michael Merideth 8

Write Operation Complete

Positive responses from quorum means write complete

Page 9: TRUST Spring Conference, April 2-3, 2008 Write Markers for Probabilistic Quorum Systems Michael Merideth, Carnegie Mellon University Michael Reiter, University

4/3/08 Michael Merideth 9

Write Operation Complete

Page 10: TRUST Spring Conference, April 2-3, 2008 Write Markers for Probabilistic Quorum Systems Michael Merideth, Carnegie Mellon University Michael Reiter, University

4/3/08 Michael Merideth 10

Read Operation

Client queries read quorum for values

Page 11: TRUST Spring Conference, April 2-3, 2008 Write Markers for Probabilistic Quorum Systems Michael Merideth, Carnegie Mellon University Michael Reiter, University

4/3/08 Michael Merideth 11

Read Operation

Determines read value based on votes (responses) from entire quorum

(Chooses “ice cream”)

Page 12: TRUST Spring Conference, April 2-3, 2008 Write Markers for Probabilistic Quorum Systems Michael Merideth, Carnegie Mellon University Michael Reiter, University

4/3/08 Michael Merideth 12

Write Markers Concept

Write marker: additional data (written with value) that identifies write quorum– Verified by clients during read

Improves properties of probabilistic quorum systems– Tolerate more faults and use smaller quorums

Page 13: TRUST Spring Conference, April 2-3, 2008 Write Markers for Probabilistic Quorum Systems Michael Merideth, Carnegie Mellon University Michael Reiter, University

4/3/08 Michael Merideth 13

Outline

Strict, Byzantine quorum systems Probabilistic, Byzantine quorum systems Benefits of write markers Idea for implementation

Page 14: TRUST Spring Conference, April 2-3, 2008 Write Markers for Probabilistic Quorum Systems Michael Merideth, Carnegie Mellon University Michael Reiter, University

4/3/08 Michael Merideth 14

Byzantine Quorum System[malkhi & reiter 98]

Byzantine (arbitrary) faults– Faulty nodes may lie– Faulty clients and servers may

collude b faulty servers

– Identity of faulty nodes unknown by non-faulty nodes

Page 15: TRUST Spring Conference, April 2-3, 2008 Write Markers for Probabilistic Quorum Systems Michael Merideth, Carnegie Mellon University Michael Reiter, University

4/3/08 Michael Merideth 15

Write Operation

Write quorum may contain faulty servers

Page 16: TRUST Spring Conference, April 2-3, 2008 Write Markers for Probabilistic Quorum Systems Michael Merideth, Carnegie Mellon University Michael Reiter, University

4/3/08 Michael Merideth 16

Write Operation Complete

Page 17: TRUST Spring Conference, April 2-3, 2008 Write Markers for Probabilistic Quorum Systems Michael Merideth, Carnegie Mellon University Michael Reiter, University

4/3/08 Michael Merideth 17

Read Operation

Faulty servers may fabricate value

Page 18: TRUST Spring Conference, April 2-3, 2008 Write Markers for Probabilistic Quorum Systems Michael Merideth, Carnegie Mellon University Michael Reiter, University

4/3/08 Michael Merideth 18

Stale Values

Stale (logically older) values are detectible

Page 19: TRUST Spring Conference, April 2-3, 2008 Write Markers for Probabilistic Quorum Systems Michael Merideth, Carnegie Mellon University Michael Reiter, University

4/3/08 Michael Merideth 19

Conflicting Values

Faulty servers may also fabricate conflicting (logically concurrent) values– E.g., same timestamp

Here “fish” conflicts with “ice cream”– But ice cream has more votes

Page 20: TRUST Spring Conference, April 2-3, 2008 Write Markers for Probabilistic Quorum Systems Michael Merideth, Carnegie Mellon University Michael Reiter, University

4/3/08 Michael Merideth 20

More Conflicting Values

Non-faulty servers may also return conflicting values

For example, in single-round write protocols– Such protocols are desirable for

efficiency– Client may (perhaps unknowingly)

submit a write that is conflicting

Page 21: TRUST Spring Conference, April 2-3, 2008 Write Markers for Probabilistic Quorum Systems Michael Merideth, Carnegie Mellon University Michael Reiter, University

4/3/08 Michael Merideth 21

Conflicting Write

Same as normal write

Page 22: TRUST Spring Conference, April 2-3, 2008 Write Markers for Probabilistic Quorum Systems Michael Merideth, Carnegie Mellon University Michael Reiter, University

4/3/08 Michael Merideth 22

Conflicting Write Incomplete

Accepted by non-faulty servers that have not accepted (conflicting) value

Write does not complete

Page 23: TRUST Spring Conference, April 2-3, 2008 Write Markers for Probabilistic Quorum Systems Michael Merideth, Carnegie Mellon University Michael Reiter, University

4/3/08 Michael Merideth 23

Which Value is Correct?

“Ice cream” was complete– … therefore is correct

“Fish” was incomplete– … therefore should be ignored

But ice cream and fish get equal votes

Client uncertain?

Page 24: TRUST Spring Conference, April 2-3, 2008 Write Markers for Probabilistic Quorum Systems Michael Merideth, Carnegie Mellon University Michael Reiter, University

4/3/08 Michael Merideth 24

Conflicting Values: Problematic

Must outvote conflicting replicas Thus, many potentially conflicting

replicas implies ability to tolerate (relatively) few faults

?

Page 25: TRUST Spring Conference, April 2-3, 2008 Write Markers for Probabilistic Quorum Systems Michael Merideth, Carnegie Mellon University Michael Reiter, University

4/3/08 Michael Merideth 25

Impact of Conflicting Replicas

Quorum Conflict Faults Protocols

Opaque< n/5 (least)

e.g., Q/U

Masking < n/4e.g., Fleet,

PASIS

Dissemin-ation

< n/3 (most)

e.g., BFT, HQ

?

Page 26: TRUST Spring Conference, April 2-3, 2008 Write Markers for Probabilistic Quorum Systems Michael Merideth, Carnegie Mellon University Michael Reiter, University

4/3/08 Michael Merideth 26

Choice of Quorums Important

Choices of read quorum and both write quorums led to problem– Other choices lead to correct

answer

?

Page 27: TRUST Spring Conference, April 2-3, 2008 Write Markers for Probabilistic Quorum Systems Michael Merideth, Carnegie Mellon University Michael Reiter, University

4/3/08 Michael Merideth 27

Choice of Quorums Important

Choices of read quorum and both write quorums led to problem– Other choices lead to correct

answer

Page 28: TRUST Spring Conference, April 2-3, 2008 Write Markers for Probabilistic Quorum Systems Michael Merideth, Carnegie Mellon University Michael Reiter, University

4/3/08 Michael Merideth 28

Idea: Select Quorums at Random

In fact, correct answer in expectation (in this example)– If quorums chosen uniformly at

random(an access strategy)

Page 29: TRUST Spring Conference, April 2-3, 2008 Write Markers for Probabilistic Quorum Systems Michael Merideth, Carnegie Mellon University Michael Reiter, University

4/3/08 Michael Merideth 29

Probabilistic Quorum Systems[malkhi, reiter, wool, wright 01]

Weakening intersection property to hold only with high probability – Provides better availability – Tolerates more faults

Bounds error probability– Probability that quorums chosen

according to access strategy yield incorrect (or uncertain) result

Page 30: TRUST Spring Conference, April 2-3, 2008 Write Markers for Probabilistic Quorum Systems Michael Merideth, Carnegie Mellon University Michael Reiter, University

4/3/08 Michael Merideth 30

Probabilistic Opaque Quorum Systems[merideth & reiter 07]

Generalize access strategy– Quorums chosen from access sets– Access sets are chosen according to

access strategy

Tolerate Byzantine clients for all probabilistic quorum systems– Enforce access strategy

Page 31: TRUST Spring Conference, April 2-3, 2008 Write Markers for Probabilistic Quorum Systems Michael Merideth, Carnegie Mellon University Michael Reiter, University

4/3/08 Michael Merideth 31

Probabilistic Quorum Systems

Reduce number of conflicting values in expectation– Therefore, tolerate more faults (with some bounded probability of error)

Conflicting

Faults

Strict Prob.

Opaque< n/5

(fewest)

Masking < n/4

Dissemination < n/3

< n/3.15

< n/2.62

< n (most)

Page 32: TRUST Spring Conference, April 2-3, 2008 Write Markers for Probabilistic Quorum Systems Michael Merideth, Carnegie Mellon University Michael Reiter, University

4/3/08 Michael Merideth 32

Reduce conflicting replicas further?

Yes (for probabilistic masking and opaque quorum systems)– Write markers

Page 33: TRUST Spring Conference, April 2-3, 2008 Write Markers for Probabilistic Quorum Systems Michael Merideth, Carnegie Mellon University Michael Reiter, University

4/3/08 Michael Merideth 33

Write Markers

Recall,– Write operations write values– Read operations poll replicas for values

Write marker– Additional data (written with value) that identifies

the write quorum (or access set) that was used– Client accepts vote (during read) only if replica was

part of write quorum (or access set)

Page 34: TRUST Spring Conference, April 2-3, 2008 Write Markers for Probabilistic Quorum Systems Michael Merideth, Carnegie Mellon University Michael Reiter, University

4/3/08 Michael Merideth 34

Write Operations with Write Markers

Create write marker for quorum

Page 35: TRUST Spring Conference, April 2-3, 2008 Write Markers for Probabilistic Quorum Systems Michael Merideth, Carnegie Mellon University Michael Reiter, University

4/3/08 Michael Merideth 35

Write Operation Complete

Page 36: TRUST Spring Conference, April 2-3, 2008 Write Markers for Probabilistic Quorum Systems Michael Merideth, Carnegie Mellon University Michael Reiter, University

4/3/08 Michael Merideth 36

Conflicting Write with Write Markers

Same as normal write

Page 37: TRUST Spring Conference, April 2-3, 2008 Write Markers for Probabilistic Quorum Systems Michael Merideth, Carnegie Mellon University Michael Reiter, University

4/3/08 Michael Merideth 37

Conflicting Write Incomplete

Accepted by non-faulty servers that have not accepted (conflicting) value

Page 38: TRUST Spring Conference, April 2-3, 2008 Write Markers for Probabilistic Quorum Systems Michael Merideth, Carnegie Mellon University Michael Reiter, University

4/3/08 Michael Merideth 38

Which Value is Correct?

“Ice cream” was complete– … therefore is correct

“Fish” was incomplete– … therefore should be ignored

Page 39: TRUST Spring Conference, April 2-3, 2008 Write Markers for Probabilistic Quorum Systems Michael Merideth, Carnegie Mellon University Michael Reiter, University

4/3/08 Michael Merideth 39

Which Value is Correct?

Faulty client can only vote for “triangle”

Faulty client cannot vote for “star”

Page 40: TRUST Spring Conference, April 2-3, 2008 Write Markers for Probabilistic Quorum Systems Michael Merideth, Carnegie Mellon University Michael Reiter, University

4/3/08 Michael Merideth 40

Benefit of Write Markers

Faulty servers cannot vote for conflicting value unless they are part of write

Due to probabilistic access strategy, faulty server not always part of write

Thus, fewer conflicting servers to outvote in expectation

Page 41: TRUST Spring Conference, April 2-3, 2008 Write Markers for Probabilistic Quorum Systems Michael Merideth, Carnegie Mellon University Michael Reiter, University

4/3/08 Michael Merideth 41

Benefits of Write Markers

Conflicting

Faults

Strict Prob.Write-

markers

Opaque< n/5 (fewest)

Masking < n/4

Dissemination < n/3

< n/3.15

< n/2.62

< n (most)

< n/2.62

< n/2

< n(most)

Tolerate more faults

Page 42: TRUST Spring Conference, April 2-3, 2008 Write Markers for Probabilistic Quorum Systems Michael Merideth, Carnegie Mellon University Michael Reiter, University

4/3/08 Michael Merideth 42

Benefits of Write Markers

Tolerate more faults Use smaller quorums

– See paper

Page 43: TRUST Spring Conference, April 2-3, 2008 Write Markers for Probabilistic Quorum Systems Michael Merideth, Carnegie Mellon University Michael Reiter, University

4/3/08 Michael Merideth 43

Example with Benign Clients

For writes: clients choose access sets uniformly at random – Then encode and, e.g., digitally sign their choices (i.e.,

create a write marker) For reads: clients verify write marker

Page 44: TRUST Spring Conference, April 2-3, 2008 Write Markers for Probabilistic Quorum Systems Michael Merideth, Carnegie Mellon University Michael Reiter, University

4/3/08 Michael Merideth 44

Write Markers with Byzantine Clients

Faulty clients:– Cannot be trusted to follow access strategy– May intentionally choose quorums that maximize

conflicting values

Constrain clients [merideth&reiter 07]– Even faulty clients follow access strategy– Avoids additional communication on critical path– Choice is verified by servers as (pseudo) random

Treat choice as write marker– Modify protocol so that clients also verify choice

Page 45: TRUST Spring Conference, April 2-3, 2008 Write Markers for Probabilistic Quorum Systems Michael Merideth, Carnegie Mellon University Michael Reiter, University

4/3/08 Michael Merideth 45

Protocol Intuition

Servers provide pseudorandom sequence of access sets per client– Threshold signature from servers

Page 46: TRUST Spring Conference, April 2-3, 2008 Write Markers for Probabilistic Quorum Systems Michael Merideth, Carnegie Mellon University Michael Reiter, University

4/3/08 Michael Merideth 46

Servers provide pseudorandom sequence of access sets per client– Threshold signature from servers

For each operation, client locally chooses next access set in sequence; servers verify choice

Protocol Intuition

Page 47: TRUST Spring Conference, April 2-3, 2008 Write Markers for Probabilistic Quorum Systems Michael Merideth, Carnegie Mellon University Michael Reiter, University

4/3/08 Michael Merideth 47

Protocol Intuition

Servers provide pseudorandom sequence of access sets per client– Threshold signature from servers

For each operation, client locally chooses next access set in sequence; servers verify choice

Page 48: TRUST Spring Conference, April 2-3, 2008 Write Markers for Probabilistic Quorum Systems Michael Merideth, Carnegie Mellon University Michael Reiter, University

4/3/08 Michael Merideth 48

Misuse by Faulty Client

What if faulty client:– Skips ahead to “better” access

set?– Waits to perform operation until

advantageous?

In either case, access set no longer random

Page 49: TRUST Spring Conference, April 2-3, 2008 Write Markers for Probabilistic Quorum Systems Michael Merideth, Carnegie Mellon University Michael Reiter, University

4/3/08 Michael Merideth 49

Defending Against Misuse

Exponential increase in cost to use later access sets– Client puzzle (requires solution)

Correct value propagates in background [c.f. malkhi et al. 03]

Sequence becomes invalid as system progresses– Must obtain new sequence

Page 50: TRUST Spring Conference, April 2-3, 2008 Write Markers for Probabilistic Quorum Systems Michael Merideth, Carnegie Mellon University Michael Reiter, University

4/3/08 Michael Merideth 50

Write Markers Mechanism

Use client puzzle – Servers already verify solution

Have clients verify as well – Treat solution and access set

as write marker– Return during read operations

Provides mechanism for write markers

Page 51: TRUST Spring Conference, April 2-3, 2008 Write Markers for Probabilistic Quorum Systems Michael Merideth, Carnegie Mellon University Michael Reiter, University

4/3/08 Michael Merideth 51

Conclusion

Write markers provide benefits for probabilistic quorum systems– Reduce number of faulty servers that can vote for

conflicting value in expectation– Increase number of faults that can be tolerated

Opaque: up to n/2.62 (probabilistic: n/3.15; strict: n/5) Masking: up to n/2 (probabilistic: n/2.62; strict: n/4)

– Allow for smaller quorums in some cases

For more information:– Write Markers for Probabilistic Quorum Systems.

Michael G. Merideth and Michael K. Reiter. CMU Technical Report: CMU-ISR-08-110

Page 52: TRUST Spring Conference, April 2-3, 2008 Write Markers for Probabilistic Quorum Systems Michael Merideth, Carnegie Mellon University Michael Reiter, University

4/3/08 Michael Merideth 52

Questions?

Page 53: TRUST Spring Conference, April 2-3, 2008 Write Markers for Probabilistic Quorum Systems Michael Merideth, Carnegie Mellon University Michael Reiter, University

4/3/08 Michael Merideth 53

Page 54: TRUST Spring Conference, April 2-3, 2008 Write Markers for Probabilistic Quorum Systems Michael Merideth, Carnegie Mellon University Michael Reiter, University

4/3/08 Michael Merideth 54

Page 55: TRUST Spring Conference, April 2-3, 2008 Write Markers for Probabilistic Quorum Systems Michael Merideth, Carnegie Mellon University Michael Reiter, University

4/3/08 Michael Merideth 55

Page 56: TRUST Spring Conference, April 2-3, 2008 Write Markers for Probabilistic Quorum Systems Michael Merideth, Carnegie Mellon University Michael Reiter, University

4/3/08 Michael Merideth 56

Page 57: TRUST Spring Conference, April 2-3, 2008 Write Markers for Probabilistic Quorum Systems Michael Merideth, Carnegie Mellon University Michael Reiter, University

4/3/08 Michael Merideth 57