Xia Zhou*, Stratis Ioannidis ♯, and Laurent Massoulié + * University of California, Santa Barbara...

28
On the Stability and Optimality of Universal Swarms Xia Zhou*, Stratis Ioannidis , and Laurent Massoulié + *University of California, Santa Barbara Technicolor Research Lab, Palo Alto + Technicolor Research Lab, Paris

Transcript of Xia Zhou*, Stratis Ioannidis ♯, and Laurent Massoulié + * University of California, Santa Barbara...

Page 1: Xia Zhou*, Stratis Ioannidis ♯, and Laurent Massoulié + * University of California, Santa Barbara ♯ Technicolor Research Lab, Palo Alto + Technicolor Research.

On the Stability and Optimality of Universal Swarms

Xia Zhou*, Stratis Ioannidis♯, and Laurent Massoulié+

*University of California, Santa Barbara♯Technicolor Research Lab, Palo Alto

+Technicolor Research Lab, Paris

Page 2: Xia Zhou*, Stratis Ioannidis ♯, and Laurent Massoulié + * University of California, Santa Barbara ♯ Technicolor Research Lab, Palo Alto + Technicolor Research.

2

• Swarm: set of users interested in the same file

Bit-Torrent Swarms

Seed

Page 3: Xia Zhou*, Stratis Ioannidis ♯, and Laurent Massoulié + * University of California, Santa Barbara ♯ Technicolor Research Lab, Palo Alto + Technicolor Research.

3

[Hajek and Zhu 10]• Unstable when λ > s!• Missing-piece syndrome: Each peer waiting for only one

piece

Bandwidth Under-Utilization

Seed

λ peers per sec

s chunks per sec

Online P2P Networks

Page 4: Xia Zhou*, Stratis Ioannidis ♯, and Laurent Massoulié + * University of California, Santa Barbara ♯ Technicolor Research Lab, Palo Alto + Technicolor Research.

4

Bandwidth Under-Utilization

• Cached content is shared in a P2P fashion (eg. bluetooth)• Opportunistic communication• May not encounter the content they are interested in

Mobile P2P Networks

?

Page 5: Xia Zhou*, Stratis Ioannidis ♯, and Laurent Massoulié + * University of California, Santa Barbara ♯ Technicolor Research Lab, Palo Alto + Technicolor Research.

5

Universal Swarms

Key idea: Exchange chunks across swarms upon bandwidth under-utilization

Question 1: How does such inter-swarm exchange affect stability?

Question 2: How should items be exchanged among swarms?

Page 6: Xia Zhou*, Stratis Ioannidis ♯, and Laurent Massoulié + * University of California, Santa Barbara ♯ Technicolor Research Lab, Palo Alto + Technicolor Research.

6

• A versatile model for universal swarms

• Universal swarms achieve better stability compared to autonomous swarms

• Only one swarm can become unstable!

• Optimal replication ratios that minimize the time for peers to retrieve interested content

Our Contributions

Page 7: Xia Zhou*, Stratis Ioannidis ♯, and Laurent Massoulié + * University of California, Santa Barbara ♯ Technicolor Research Lab, Palo Alto + Technicolor Research.

7

• Motivation

• A model for universal swarms

• Main results• Stability of universal swarms• Content exchange designs in universal swarms

• Conclusion and future works

Outline

Page 8: Xia Zhou*, Stratis Ioannidis ♯, and Laurent Massoulié + * University of California, Santa Barbara ♯ Technicolor Research Lab, Palo Alto + Technicolor Research.

8

• Peer requests one chunk i K• Peers requesting the same chunk form a peer

swarm

Peer Swarms

??

?

Page 9: Xia Zhou*, Stratis Ioannidis ♯, and Laurent Massoulié + * University of California, Santa Barbara ♯ Technicolor Research Lab, Palo Alto + Technicolor Research.

9

• Peer has cache size of C• Peer may use cache to store chunks it is not

interested in

Peer Caches

Cache

?

Request

C

Stored chunks f K

Page 10: Xia Zhou*, Stratis Ioannidis ♯, and Laurent Massoulié + * University of California, Santa Barbara ♯ Technicolor Research Lab, Palo Alto + Technicolor Research.

10

• Peers arrive with full caches• Peers requesting i and caching f arrive

according to a Poisson process with rate λi, f

Peer Arrivals

Time

? ? ?Cache Reques

t

C ?

Page 11: Xia Zhou*, Stratis Ioannidis ♯, and Laurent Massoulié + * University of California, Santa Barbara ♯ Technicolor Research Lab, Palo Alto + Technicolor Research.

11

• Online P2P: random sampling• Mobile P2P: contact when within transmission range

Peer Contact Process

Time

?

×

? ?

×

Page 12: Xia Zhou*, Stratis Ioannidis ♯, and Laurent Massoulié + * University of California, Santa Barbara ♯ Technicolor Research Lab, Palo Alto + Technicolor Research.

• One peer contacts other peers according to a Poisson process with rate

• N(t): number of peers in the system at time t

12

Peer Contact Process (Cont.)

N(t)

Contact rate

0 ≤ β < 1

Constant-bandwidth

N(t)

Contact rate

μ

β = 1

N(t)

Contact rate

1 < β ≤ 2

Contact-constrained

Interference-constrained

Page 13: Xia Zhou*, Stratis Ioannidis ♯, and Laurent Massoulié + * University of California, Santa Barbara ♯ Technicolor Research Lab, Palo Alto + Technicolor Research.

13

• If encountering requested chunk: Grab-and-Go• Otherwise:• Static-cache policy: no change on cached chunks• Alternatives: updating cached contents

• Requested chunk and cached chunks define a peer class• N(t): system state at time t (# of peers in each peer class)

Content Exchange Policy

? ??

Conversion probability

A, A’ B, B’

Page 14: Xia Zhou*, Stratis Ioannidis ♯, and Laurent Massoulié + * University of California, Santa Barbara ♯ Technicolor Research Lab, Palo Alto + Technicolor Research.

14

• Motivation

• Model for universal swarms

• Main results• Stability of universal swarms• Content exchange designs in universal swarms

• Conclusion and future works

Outline

Page 15: Xia Zhou*, Stratis Ioannidis ♯, and Laurent Massoulié + * University of California, Santa Barbara ♯ Technicolor Research Lab, Palo Alto + Technicolor Research.

15

The evolution of the universal swarm system can be approximated arbitrarily well by the solution of a system of ODEs that depend on the conversion probabilities

Methodology: Fluid Limit

• For all β

• For all content exchange policies

Page 16: Xia Zhou*, Stratis Ioannidis ♯, and Laurent Massoulié + * University of California, Santa Barbara ♯ Technicolor Research Lab, Palo Alto + Technicolor Research.

16

Question 1:

How does inter-swarm exchange affect the system stability?

Universal Swarms

Page 17: Xia Zhou*, Stratis Ioannidis ♯, and Laurent Massoulié + * University of California, Santa Barbara ♯ Technicolor Research Lab, Palo Alto + Technicolor Research.

17

Stability of Static-Cache PolicyLet > 0 be the arrival rate of peers requesting i and storing j.

17

Theorem: The system is stable under the static cache policy if and only if:

• Independent of β and cache size C

• The system is stable even if arrivals of peers requesting i exceed arrivals of peers storing i!

Page 18: Xia Zhou*, Stratis Ioannidis ♯, and Laurent Massoulié + * University of California, Santa Barbara ♯ Technicolor Research Lab, Palo Alto + Technicolor Research.

18

Only One Swarm Can Become Unstable!

• At most one swarm can blow up!

?

?

?

Page 19: Xia Zhou*, Stratis Ioannidis ♯, and Laurent Massoulié + * University of California, Santa Barbara ♯ Technicolor Research Lab, Palo Alto + Technicolor Research.

19

• Motivation

• Model for universal swarms

• Main results• Stability of universal swarms• Content exchange designs in universal

swarms

• Conclusion and future works

Outline

Page 20: Xia Zhou*, Stratis Ioannidis ♯, and Laurent Massoulié + * University of California, Santa Barbara ♯ Technicolor Research Lab, Palo Alto + Technicolor Research.

20

Question 2:

How should chunks be exchanged across swarms?

Universal Swarms

Page 21: Xia Zhou*, Stratis Ioannidis ♯, and Laurent Massoulié + * University of California, Santa Barbara ♯ Technicolor Research Lab, Palo Alto + Technicolor Research.

21

Optimal Demand and Supply

-- the number of peers requesting chunk i (demand)

-- the number of peers storing chunk i (supply)

Theorem: Under the grab-and-go principle, the average sojourn time of a peer in the system is minimized when

where .

• The optimal supply is C times the demand!

Page 22: Xia Zhou*, Stratis Ioannidis ♯, and Laurent Massoulié + * University of California, Santa Barbara ♯ Technicolor Research Lab, Palo Alto + Technicolor Research.

22

• Centralized tracker maintains valuation vi for each chunk i

• Positive vi : chunk i needs more replicas

• Negative vi : chunk i needs fewer replicas

• Replace the chunks with negative valuation with that with positive valuations

BARON: Valuation-Guided Replication

2 1 0 -1

??

Page 23: Xia Zhou*, Stratis Ioannidis ♯, and Laurent Massoulié + * University of California, Santa Barbara ♯ Technicolor Research Lab, Palo Alto + Technicolor Research.

23

-- the number of peers requesting chunk i in the optimal state

-- the number of peers storing chunk i in the optimal state

BARON: Valuation Design

Optimal:

Valuation:

No need to know arrival rates and contact rates, but only the cache size C

Need to track the demand and supply dynamically

Page 24: Xia Zhou*, Stratis Ioannidis ♯, and Laurent Massoulié + * University of California, Santa Barbara ♯ Technicolor Research Lab, Palo Alto + Technicolor Research.

24

• Evaluations based on fluid trajectories in MATLAB

• Numerically solving ODEs

BARON: Numerical Results

Valuation-guided content exchange improves the system stability

Page 25: Xia Zhou*, Stratis Ioannidis ♯, and Laurent Massoulié + * University of California, Santa Barbara ♯ Technicolor Research Lab, Palo Alto + Technicolor Research.

25

• Universal swarms achieve better stability even with the simplest replication strategy

• At most one swarm can blow up!

• Optimal supply linearly proportional to the demand

• BARON extends the stability region using valuations

• Better understanding of the dynamics under more sophisticated content exchange mechanisms

• Peer incentives• Removing the assumption of one-chunk request

Conclusion and Future Works

Page 26: Xia Zhou*, Stratis Ioannidis ♯, and Laurent Massoulié + * University of California, Santa Barbara ♯ Technicolor Research Lab, Palo Alto + Technicolor Research.

Thank you!

Page 27: Xia Zhou*, Stratis Ioannidis ♯, and Laurent Massoulié + * University of California, Santa Barbara ♯ Technicolor Research Lab, Palo Alto + Technicolor Research.

Backup

Page 28: Xia Zhou*, Stratis Ioannidis ♯, and Laurent Massoulié + * University of California, Santa Barbara ♯ Technicolor Research Lab, Palo Alto + Technicolor Research.

28

Only One Swarm Can Become Unstable!

Let > 0 be the arrival rate of peers requesting i and storing j.

Theorem: There exists at most one item i for which

Moreover, for β in [0,1], the number of peers requesting item i grows to infinity, while the number of peers requesting other items remains bounded.

• At most one swarm can blow up!