university of michigan near-optimal latency versus cost mosharaf chowdhury, harsha … ·...
TRANSCRIPT
![Page 1: University of Michigan Near-Optimal Latency Versus Cost Mosharaf Chowdhury, Harsha … · 2020-02-28 · Mosharaf Chowdhury, Harsha V. Madhyastha University of Michigan. Web Ser ver](https://reader033.vdocuments.us/reader033/viewer/2022042413/5f2d85ced7747d45783ae8ba/html5/thumbnails/1.jpg)
Near-Optimal Latency Versus Cost Tradeoffs in Geo-Distributed StorageMuhammed Uluyol, Anthony Huang, Ayush Goel,Mosharaf Chowdhury, Harsha V. Madhyastha
University of Michigan
![Page 2: University of Michigan Near-Optimal Latency Versus Cost Mosharaf Chowdhury, Harsha … · 2020-02-28 · Mosharaf Chowdhury, Harsha V. Madhyastha University of Michigan. Web Ser ver](https://reader033.vdocuments.us/reader033/viewer/2022042413/5f2d85ced7747d45783ae8ba/html5/thumbnails/2.jpg)
Web Server
Data Site
Distribute Web Servers for Interactive Latency
2
![Page 3: University of Michigan Near-Optimal Latency Versus Cost Mosharaf Chowdhury, Harsha … · 2020-02-28 · Mosharaf Chowdhury, Harsha V. Madhyastha University of Michigan. Web Ser ver](https://reader033.vdocuments.us/reader033/viewer/2022042413/5f2d85ced7747d45783ae8ba/html5/thumbnails/3.jpg)
Data Site
Distribute Data for Availability
3
Web Server
![Page 4: University of Michigan Near-Optimal Latency Versus Cost Mosharaf Chowdhury, Harsha … · 2020-02-28 · Mosharaf Chowdhury, Harsha V. Madhyastha University of Michigan. Web Ser ver](https://reader033.vdocuments.us/reader033/viewer/2022042413/5f2d85ced7747d45783ae8ba/html5/thumbnails/4.jpg)
Data Site
Distribute Data for Availability and Latency
4
Web Server
![Page 5: University of Michigan Near-Optimal Latency Versus Cost Mosharaf Chowdhury, Harsha … · 2020-02-28 · Mosharaf Chowdhury, Harsha V. Madhyastha University of Michigan. Web Ser ver](https://reader033.vdocuments.us/reader033/viewer/2022042413/5f2d85ced7747d45783ae8ba/html5/thumbnails/5.jpg)
Linearizability Imposes Unavoidable Trade-offs
5
● Read vs Write Latency
● Read Latency vs Cost
![Page 6: University of Michigan Near-Optimal Latency Versus Cost Mosharaf Chowdhury, Harsha … · 2020-02-28 · Mosharaf Chowdhury, Harsha V. Madhyastha University of Michigan. Web Ser ver](https://reader033.vdocuments.us/reader033/viewer/2022042413/5f2d85ced7747d45783ae8ba/html5/thumbnails/6.jpg)
Linearizability Imposes Unavoidable Trade-offs
6
● Read vs Write Latency
● Read Latency vs Cost
Read
Overlap ensures consistency
Write
![Page 7: University of Michigan Near-Optimal Latency Versus Cost Mosharaf Chowdhury, Harsha … · 2020-02-28 · Mosharaf Chowdhury, Harsha V. Madhyastha University of Michigan. Web Ser ver](https://reader033.vdocuments.us/reader033/viewer/2022042413/5f2d85ced7747d45783ae8ba/html5/thumbnails/7.jpg)
Linearizability Imposes Unavoidable Trade-offs
7
● Read vs Write Latency
● Read Latency vs Cost
Read
Overlap ensures consistency
Write
![Page 8: University of Michigan Near-Optimal Latency Versus Cost Mosharaf Chowdhury, Harsha … · 2020-02-28 · Mosharaf Chowdhury, Harsha V. Madhyastha University of Michigan. Web Ser ver](https://reader033.vdocuments.us/reader033/viewer/2022042413/5f2d85ced7747d45783ae8ba/html5/thumbnails/8.jpg)
Linearizability Imposes Unavoidable Trade-offs
8
● Read vs Write Latency
● Read Latency vs Cost
Read
Write
![Page 9: University of Michigan Near-Optimal Latency Versus Cost Mosharaf Chowdhury, Harsha … · 2020-02-28 · Mosharaf Chowdhury, Harsha V. Madhyastha University of Michigan. Web Ser ver](https://reader033.vdocuments.us/reader033/viewer/2022042413/5f2d85ced7747d45783ae8ba/html5/thumbnails/9.jpg)
Linearizability Imposes Unavoidable Trade-offs
9
● Read vs Write Latency
● Read Latency vs Cost
![Page 10: University of Michigan Near-Optimal Latency Versus Cost Mosharaf Chowdhury, Harsha … · 2020-02-28 · Mosharaf Chowdhury, Harsha V. Madhyastha University of Michigan. Web Ser ver](https://reader033.vdocuments.us/reader033/viewer/2022042413/5f2d85ced7747d45783ae8ba/html5/thumbnails/10.jpg)
Linearizability Imposes Unavoidable Trade-offs
10
● Read vs Write Latency
● Read Latency vs Cost
![Page 11: University of Michigan Near-Optimal Latency Versus Cost Mosharaf Chowdhury, Harsha … · 2020-02-28 · Mosharaf Chowdhury, Harsha V. Madhyastha University of Michigan. Web Ser ver](https://reader033.vdocuments.us/reader033/viewer/2022042413/5f2d85ced7747d45783ae8ba/html5/thumbnails/11.jpg)
Linearizability Imposes Unavoidable Trade-offs
11
● Read vs Write Latency
● Read Latency vs Cost
![Page 12: University of Michigan Near-Optimal Latency Versus Cost Mosharaf Chowdhury, Harsha … · 2020-02-28 · Mosharaf Chowdhury, Harsha V. Madhyastha University of Michigan. Web Ser ver](https://reader033.vdocuments.us/reader033/viewer/2022042413/5f2d85ced7747d45783ae8ba/html5/thumbnails/12.jpg)
Linearizability Imposes Unavoidable Trade-offs
12
● Read vs Write Latency
● Read Latency vs Cost
![Page 13: University of Michigan Near-Optimal Latency Versus Cost Mosharaf Chowdhury, Harsha … · 2020-02-28 · Mosharaf Chowdhury, Harsha V. Madhyastha University of Michigan. Web Ser ver](https://reader033.vdocuments.us/reader033/viewer/2022042413/5f2d85ced7747d45783ae8ba/html5/thumbnails/13.jpg)
How Do Existing Approaches Perform?
13
Low
est P
ossi
ble
Rea
d La
tenc
y
StorageOverhead Budget
Write Latency Budget
0
![Page 14: University of Michigan Near-Optimal Latency Versus Cost Mosharaf Chowdhury, Harsha … · 2020-02-28 · Mosharaf Chowdhury, Harsha V. Madhyastha University of Michigan. Web Ser ver](https://reader033.vdocuments.us/reader033/viewer/2022042413/5f2d85ced7747d45783ae8ba/html5/thumbnails/14.jpg)
How Do Existing Approaches Perform?
14
Better
0
Low
est P
ossi
ble
Rea
d La
tenc
y
Write Latency BudgetStorage
Overhead Budget
![Page 15: University of Michigan Near-Optimal Latency Versus Cost Mosharaf Chowdhury, Harsha … · 2020-02-28 · Mosharaf Chowdhury, Harsha V. Madhyastha University of Michigan. Web Ser ver](https://reader033.vdocuments.us/reader033/viewer/2022042413/5f2d85ced7747d45783ae8ba/html5/thumbnails/15.jpg)
How Do Existing Approaches Perform?
15
0
Low
est P
ossi
ble
Rea
d La
tenc
y
Write Latency BudgetStorage
Overhead Budget
Better
![Page 16: University of Michigan Near-Optimal Latency Versus Cost Mosharaf Chowdhury, Harsha … · 2020-02-28 · Mosharaf Chowdhury, Harsha V. Madhyastha University of Michigan. Web Ser ver](https://reader033.vdocuments.us/reader033/viewer/2022042413/5f2d85ced7747d45783ae8ba/html5/thumbnails/16.jpg)
How Do Existing Approaches Perform?
16
Better
Better
0
Low
est P
ossi
ble
Rea
d La
tenc
y
Write Latency BudgetStorage
Overhead Budget
![Page 17: University of Michigan Near-Optimal Latency Versus Cost Mosharaf Chowdhury, Harsha … · 2020-02-28 · Mosharaf Chowdhury, Harsha V. Madhyastha University of Michigan. Web Ser ver](https://reader033.vdocuments.us/reader033/viewer/2022042413/5f2d85ced7747d45783ae8ba/html5/thumbnails/17.jpg)
● EPaxos: state-of-the-art geo-replication protocol
How Do Existing Approaches Perform?
17
![Page 18: University of Michigan Near-Optimal Latency Versus Cost Mosharaf Chowdhury, Harsha … · 2020-02-28 · Mosharaf Chowdhury, Harsha V. Madhyastha University of Michigan. Web Ser ver](https://reader033.vdocuments.us/reader033/viewer/2022042413/5f2d85ced7747d45783ae8ba/html5/thumbnails/18.jpg)
● EPaxos: state-of-the-art geo-replication protocol
● Compare with estimate of theoretical lower bound
How Do Existing Approaches Perform?
18
![Page 19: University of Michigan Near-Optimal Latency Versus Cost Mosharaf Chowdhury, Harsha … · 2020-02-28 · Mosharaf Chowdhury, Harsha V. Madhyastha University of Michigan. Web Ser ver](https://reader033.vdocuments.us/reader033/viewer/2022042413/5f2d85ced7747d45783ae8ba/html5/thumbnails/19.jpg)
● EPaxos: state-of-the-art geo-replication protocol
● Compare with estimate of theoretical lower bound○ No particular protocol○ Respects consistency and
fault-tolerance constraints
How Do Existing Approaches Perform?
19
![Page 20: University of Michigan Near-Optimal Latency Versus Cost Mosharaf Chowdhury, Harsha … · 2020-02-28 · Mosharaf Chowdhury, Harsha V. Madhyastha University of Michigan. Web Ser ver](https://reader033.vdocuments.us/reader033/viewer/2022042413/5f2d85ced7747d45783ae8ba/html5/thumbnails/20.jpg)
● EPaxos: state-of-the-art geo-replication protocol
● Compare with estimate of theoretical lower bound○ No particular protocol○ Respects consistency and
fault-tolerance constraints
How Do Existing Approaches Perform?
20
2⨉ storage forsame latency
2⨉ readlatency
![Page 21: University of Michigan Near-Optimal Latency Versus Cost Mosharaf Chowdhury, Harsha … · 2020-02-28 · Mosharaf Chowdhury, Harsha V. Madhyastha University of Michigan. Web Ser ver](https://reader033.vdocuments.us/reader033/viewer/2022042413/5f2d85ced7747d45783ae8ba/html5/thumbnails/21.jpg)
● EPaxos: state-of-the-art geo-replication protocol
● Compare with estimate of theoretical lower bound○ No particular protocol○ Respects consistency and
fault-tolerance constraints
How Do Existing Approaches Perform?
21
2⨉ storage forsame latency
2⨉ readlatency
Core problem: replicationEach site stores a full copy
![Page 22: University of Michigan Near-Optimal Latency Versus Cost Mosharaf Chowdhury, Harsha … · 2020-02-28 · Mosharaf Chowdhury, Harsha V. Madhyastha University of Michigan. Web Ser ver](https://reader033.vdocuments.us/reader033/viewer/2022042413/5f2d85ced7747d45783ae8ba/html5/thumbnails/22.jpg)
● Each site stores 1/kth of the data
● RS-Paxos: Paxos on erasure-coded data
Lowering Cost with Erasure Coding Utility of RS-Paxos
22
![Page 23: University of Michigan Near-Optimal Latency Versus Cost Mosharaf Chowdhury, Harsha … · 2020-02-28 · Mosharaf Chowdhury, Harsha V. Madhyastha University of Michigan. Web Ser ver](https://reader033.vdocuments.us/reader033/viewer/2022042413/5f2d85ced7747d45783ae8ba/html5/thumbnails/23.jpg)
● Each site stores 1/kth of the data
● RS-Paxos: Paxos on erasure-coded data
Lowering Cost with Erasure Coding
23
1.5⨉ write latency for same read
1.5⨉ latency at same storage
![Page 24: University of Michigan Near-Optimal Latency Versus Cost Mosharaf Chowdhury, Harsha … · 2020-02-28 · Mosharaf Chowdhury, Harsha V. Madhyastha University of Michigan. Web Ser ver](https://reader033.vdocuments.us/reader033/viewer/2022042413/5f2d85ced7747d45783ae8ba/html5/thumbnails/24.jpg)
● Each site stores 1/kth of the data
● RS-Paxos: Paxos on erasure-coded data
Lowering Cost with Erasure Coding
24
1.5⨉ write latency for same read
1.5⨉ latency at same storage
RS-Paxos Limitations
● Two-round writes● k-site intersection between quorums
![Page 25: University of Michigan Near-Optimal Latency Versus Cost Mosharaf Chowdhury, Harsha … · 2020-02-28 · Mosharaf Chowdhury, Harsha V. Madhyastha University of Michigan. Web Ser ver](https://reader033.vdocuments.us/reader033/viewer/2022042413/5f2d85ced7747d45783ae8ba/html5/thumbnails/25.jpg)
Recap of the Problem
● Want to spread data across DCs, but constraints that impose trade-offs
● State-of-the-art falls short of the optimal
● Use erasure coding → hurts latency
25
![Page 26: University of Michigan Near-Optimal Latency Versus Cost Mosharaf Chowdhury, Harsha … · 2020-02-28 · Mosharaf Chowdhury, Harsha V. Madhyastha University of Michigan. Web Ser ver](https://reader033.vdocuments.us/reader033/viewer/2022042413/5f2d85ced7747d45783ae8ba/html5/thumbnails/26.jpg)
● Two-round writesApproximates latency of one-round writes
● k-site intersection between quorums1-site intersection (common-case)
Pando: Near-Optimal Trade-off
26
![Page 27: University of Michigan Near-Optimal Latency Versus Cost Mosharaf Chowdhury, Harsha … · 2020-02-28 · Mosharaf Chowdhury, Harsha V. Madhyastha University of Michigan. Web Ser ver](https://reader033.vdocuments.us/reader033/viewer/2022042413/5f2d85ced7747d45783ae8ba/html5/thumbnails/27.jpg)
Paxos Review
27
![Page 28: University of Michigan Near-Optimal Latency Versus Cost Mosharaf Chowdhury, Harsha … · 2020-02-28 · Mosharaf Chowdhury, Harsha V. Madhyastha University of Michigan. Web Ser ver](https://reader033.vdocuments.us/reader033/viewer/2022042413/5f2d85ced7747d45783ae8ba/html5/thumbnails/28.jpg)
Paxos Review
● 2-Phase writes: first become leader
I am leader Ack.
30 ms 30 ms
28
![Page 29: University of Michigan Near-Optimal Latency Versus Cost Mosharaf Chowdhury, Harsha … · 2020-02-28 · Mosharaf Chowdhury, Harsha V. Madhyastha University of Michigan. Web Ser ver](https://reader033.vdocuments.us/reader033/viewer/2022042413/5f2d85ced7747d45783ae8ba/html5/thumbnails/29.jpg)
Paxos Review
● 2-Phase writes: first become leader, then write
I am leader Ack. Write
data Ack.
30 ms 30 ms 30 ms 30 ms
29
![Page 30: University of Michigan Near-Optimal Latency Versus Cost Mosharaf Chowdhury, Harsha … · 2020-02-28 · Mosharaf Chowdhury, Harsha V. Madhyastha University of Michigan. Web Ser ver](https://reader033.vdocuments.us/reader033/viewer/2022042413/5f2d85ced7747d45783ae8ba/html5/thumbnails/30.jpg)
Paxos Review
● 2-Phase writes: first become leader, then write
I am leader Ack. Write
data Ack.
30 ms 30 ms 30 ms 30 ms
30
One-round write protocol
![Page 31: University of Michigan Near-Optimal Latency Versus Cost Mosharaf Chowdhury, Harsha … · 2020-02-28 · Mosharaf Chowdhury, Harsha V. Madhyastha University of Michigan. Web Ser ver](https://reader033.vdocuments.us/reader033/viewer/2022042413/5f2d85ced7747d45783ae8ba/html5/thumbnails/31.jpg)
Quickly Executing 2-Phase Writes
● Step 1: faster Phase 1○ Flexible Paxos [OPODIS’16]: need Phase 1, 2 quorums to intersect○ Phase 1 quorums need not overlap
Write data
30 ms
Ack.
30 ms
Ack.
10 ms
I am leader
10 ms
31
![Page 32: University of Michigan Near-Optimal Latency Versus Cost Mosharaf Chowdhury, Harsha … · 2020-02-28 · Mosharaf Chowdhury, Harsha V. Madhyastha University of Michigan. Web Ser ver](https://reader033.vdocuments.us/reader033/viewer/2022042413/5f2d85ced7747d45783ae8ba/html5/thumbnails/32.jpg)
Quickly Executing 2-Phase Writes
● Step 1: faster Phase 1● Step 2: overlap latency cost of Phase 1 with Phase 2
○ RPC Chains [NSDI’09]: start Phase 2 at a delegate
32
30 ms
Ack.Write data
15 ms
Ack.
10 ms
I am leader
10 ms
![Page 33: University of Michigan Near-Optimal Latency Versus Cost Mosharaf Chowdhury, Harsha … · 2020-02-28 · Mosharaf Chowdhury, Harsha V. Madhyastha University of Michigan. Web Ser ver](https://reader033.vdocuments.us/reader033/viewer/2022042413/5f2d85ced7747d45783ae8ba/html5/thumbnails/33.jpg)
● Two-round writesApproximates latency of one-round writes
● k-site intersection between quorums1-site intersection (common-case)
Pando: Near-Optimal Trade-off
33
✔
![Page 34: University of Michigan Near-Optimal Latency Versus Cost Mosharaf Chowdhury, Harsha … · 2020-02-28 · Mosharaf Chowdhury, Harsha V. Madhyastha University of Michigan. Web Ser ver](https://reader033.vdocuments.us/reader033/viewer/2022042413/5f2d85ced7747d45783ae8ba/html5/thumbnails/34.jpg)
Write to All
Phase 2
34
k=2
![Page 35: University of Michigan Near-Optimal Latency Versus Cost Mosharaf Chowdhury, Harsha … · 2020-02-28 · Mosharaf Chowdhury, Harsha V. Madhyastha University of Michigan. Web Ser ver](https://reader033.vdocuments.us/reader033/viewer/2022042413/5f2d85ced7747d45783ae8ba/html5/thumbnails/35.jpg)
Write to All, Wait for Quorum
Phase 2
35
k=2
![Page 36: University of Michigan Near-Optimal Latency Versus Cost Mosharaf Chowdhury, Harsha … · 2020-02-28 · Mosharaf Chowdhury, Harsha V. Madhyastha University of Michigan. Web Ser ver](https://reader033.vdocuments.us/reader033/viewer/2022042413/5f2d85ced7747d45783ae8ba/html5/thumbnails/36.jpg)
v2
v2
v2
v2
Write to All, Wait for Quorum
Phase 2
36
v2
v2
v1
v2
Read
Rare Case Common Casek=2
Read
![Page 37: University of Michigan Near-Optimal Latency Versus Cost Mosharaf Chowdhury, Harsha … · 2020-02-28 · Mosharaf Chowdhury, Harsha V. Madhyastha University of Michigan. Web Ser ver](https://reader033.vdocuments.us/reader033/viewer/2022042413/5f2d85ced7747d45783ae8ba/html5/thumbnails/37.jpg)
v2
v2
v2
v2
Achieving 1-Site Intersection
Phase 2
37
v2
v2
v1
v2
Read
Rare Case Common Casek=2
Read
Maybea write finished
![Page 38: University of Michigan Near-Optimal Latency Versus Cost Mosharaf Chowdhury, Harsha … · 2020-02-28 · Mosharaf Chowdhury, Harsha V. Madhyastha University of Michigan. Web Ser ver](https://reader033.vdocuments.us/reader033/viewer/2022042413/5f2d85ced7747d45783ae8ba/html5/thumbnails/38.jpg)
v2
v2
v2
v2
Achieving 1-Site Intersection
Phase 2
38
v2
v2
v1
v2
Read
Rare Case Common Casek=2
Read
Maybea write finished
![Page 39: University of Michigan Near-Optimal Latency Versus Cost Mosharaf Chowdhury, Harsha … · 2020-02-28 · Mosharaf Chowdhury, Harsha V. Madhyastha University of Michigan. Web Ser ver](https://reader033.vdocuments.us/reader033/viewer/2022042413/5f2d85ced7747d45783ae8ba/html5/thumbnails/39.jpg)
● Two-round writesApproximates latency of one-round writes
● k-site intersection between quorums1-site intersection (common-case)
Pando: Near-Optimal Trade-off
39
✔
✔
![Page 40: University of Michigan Near-Optimal Latency Versus Cost Mosharaf Chowdhury, Harsha … · 2020-02-28 · Mosharaf Chowdhury, Harsha V. Madhyastha University of Michigan. Web Ser ver](https://reader033.vdocuments.us/reader033/viewer/2022042413/5f2d85ced7747d45783ae8ba/html5/thumbnails/40.jpg)
● Two-round writesApproximates latency of one-round writes
● k-site intersection between quorums1-site intersection (common-case)
Pando: Near-Optimal Trade-off
40
✔
✔
See paper:● Correctness● Bounding latency under conflicts
![Page 41: University of Michigan Near-Optimal Latency Versus Cost Mosharaf Chowdhury, Harsha … · 2020-02-28 · Mosharaf Chowdhury, Harsha V. Madhyastha University of Michigan. Web Ser ver](https://reader033.vdocuments.us/reader033/viewer/2022042413/5f2d85ced7747d45783ae8ba/html5/thumbnails/41.jpg)
Evaluation: Proximity to Lower Bound
● Access set: DCs hosting web servers reading/writing data
● MIP solver selects data sites to minimize latency
● 500 access sets
41
Sample access set
Measure gap
![Page 42: University of Michigan Near-Optimal Latency Versus Cost Mosharaf Chowdhury, Harsha … · 2020-02-28 · Mosharaf Chowdhury, Harsha V. Madhyastha University of Michigan. Web Ser ver](https://reader033.vdocuments.us/reader033/viewer/2022042413/5f2d85ced7747d45783ae8ba/html5/thumbnails/42.jpg)
Pando is Close to the Lower Bound
42
11⨉ higher
13⨉ higher
![Page 43: University of Michigan Near-Optimal Latency Versus Cost Mosharaf Chowdhury, Harsha … · 2020-02-28 · Mosharaf Chowdhury, Harsha V. Madhyastha University of Michigan. Web Ser ver](https://reader033.vdocuments.us/reader033/viewer/2022042413/5f2d85ced7747d45783ae8ba/html5/thumbnails/43.jpg)
Pando is Close to the Lower Bound
43
8⨉ higher
11⨉ higher
13⨉ higher
∪
![Page 44: University of Michigan Near-Optimal Latency Versus Cost Mosharaf Chowdhury, Harsha … · 2020-02-28 · Mosharaf Chowdhury, Harsha V. Madhyastha University of Michigan. Web Ser ver](https://reader033.vdocuments.us/reader033/viewer/2022042413/5f2d85ced7747d45783ae8ba/html5/thumbnails/44.jpg)
Pando is Close to the Lower Bound
44
8⨉ higher
11⨉ higher
13⨉ higherPotential for true 1-round writes
∪
![Page 45: University of Michigan Near-Optimal Latency Versus Cost Mosharaf Chowdhury, Harsha … · 2020-02-28 · Mosharaf Chowdhury, Harsha V. Madhyastha University of Michigan. Web Ser ver](https://reader033.vdocuments.us/reader033/viewer/2022042413/5f2d85ced7747d45783ae8ba/html5/thumbnails/45.jpg)
Pando is Close to the Lower Bound
45
8⨉ higher
11⨉ higher
13⨉ higherPotential for true 1-round writes
∪
● Cloud deployment confirms solver latency estimates
● Up to 46% cost ($) savings
![Page 46: University of Michigan Near-Optimal Latency Versus Cost Mosharaf Chowdhury, Harsha … · 2020-02-28 · Mosharaf Chowdhury, Harsha V. Madhyastha University of Michigan. Web Ser ver](https://reader033.vdocuments.us/reader033/viewer/2022042413/5f2d85ced7747d45783ae8ba/html5/thumbnails/46.jpg)
Conclusion
● Pando: linearizability across geo-distributed DCs
● Achieves a near-optimal read–write–storage trade-off○ Allow for erasure-code data to minimize cost○ Rethink how to use Paxos in the wide-area setting
46
![Page 47: University of Michigan Near-Optimal Latency Versus Cost Mosharaf Chowdhury, Harsha … · 2020-02-28 · Mosharaf Chowdhury, Harsha V. Madhyastha University of Michigan. Web Ser ver](https://reader033.vdocuments.us/reader033/viewer/2022042413/5f2d85ced7747d45783ae8ba/html5/thumbnails/47.jpg)
Backup Slides
47
![Page 48: University of Michigan Near-Optimal Latency Versus Cost Mosharaf Chowdhury, Harsha … · 2020-02-28 · Mosharaf Chowdhury, Harsha V. Madhyastha University of Michigan. Web Ser ver](https://reader033.vdocuments.us/reader033/viewer/2022042413/5f2d85ced7747d45783ae8ba/html5/thumbnails/48.jpg)
Deployment Latency
48
![Page 49: University of Michigan Near-Optimal Latency Versus Cost Mosharaf Chowdhury, Harsha … · 2020-02-28 · Mosharaf Chowdhury, Harsha V. Madhyastha University of Michigan. Web Ser ver](https://reader033.vdocuments.us/reader033/viewer/2022042413/5f2d85ced7747d45783ae8ba/html5/thumbnails/49.jpg)
Latency Under Conflicts
49
![Page 50: University of Michigan Near-Optimal Latency Versus Cost Mosharaf Chowdhury, Harsha … · 2020-02-28 · Mosharaf Chowdhury, Harsha V. Madhyastha University of Michigan. Web Ser ver](https://reader033.vdocuments.us/reader033/viewer/2022042413/5f2d85ced7747d45783ae8ba/html5/thumbnails/50.jpg)
Contributions of Each Technique
50
![Page 51: University of Michigan Near-Optimal Latency Versus Cost Mosharaf Chowdhury, Harsha … · 2020-02-28 · Mosharaf Chowdhury, Harsha V. Madhyastha University of Michigan. Web Ser ver](https://reader033.vdocuments.us/reader033/viewer/2022042413/5f2d85ced7747d45783ae8ba/html5/thumbnails/51.jpg)
Throughput
51
![Page 52: University of Michigan Near-Optimal Latency Versus Cost Mosharaf Chowdhury, Harsha … · 2020-02-28 · Mosharaf Chowdhury, Harsha V. Madhyastha University of Michigan. Web Ser ver](https://reader033.vdocuments.us/reader033/viewer/2022042413/5f2d85ced7747d45783ae8ba/html5/thumbnails/52.jpg)
Read Latency After Failure
52