18
Efficient Proactive Caching for Supporting Seamless Mobility Vasilios Siris, Xenofon Vasilakos, George Polyzos PhD student, AUEB, MMLAB mm.aueb.gr/~xvas [email protected]

Efficient Proactive Caching for Supporting Seamless Mobility · Efficient Proactive Caching (EPC) • Individual requests imply higher demand for cache space • Congestion pricing

  • Upload
    others

  • View
    0

  • Download
    0

Embed Size (px)

Citation preview

Page 1: Efficient Proactive Caching for Supporting Seamless Mobility · Efficient Proactive Caching (EPC) • Individual requests imply higher demand for cache space • Congestion pricing

Efficient Proactive Caching for Supporting Seamless Mobility

Vasilios Siris, Xenofon Vasilakos, George PolyzosPhD student, AUEB, MMLAB

mm.aueb.gr/[email protected]

Page 2: Efficient Proactive Caching for Supporting Seamless Mobility · Efficient Proactive Caching (EPC) • Individual requests imply higher demand for cache space • Congestion pricing

Problem

• Reduce propagation delay– f(#network hops)

Xenofon Vasilakos ‐ [email protected] 2

Page 3: Efficient Proactive Caching for Supporting Seamless Mobility · Efficient Proactive Caching (EPC) • Individual requests imply higher demand for cache space • Congestion pricing

Approach (1/2)

• Proactively fetch data‐objects to attachment points

• Is this a typical proactive caching approach?

Xenofon Vasilakos ‐ [email protected] 3

Page 4: Efficient Proactive Caching for Supporting Seamless Mobility · Efficient Proactive Caching (EPC) • Individual requests imply higher demand for cache space • Congestion pricing

Approach (2/2)

• Handoff mobility probabilities

• Exploit Individual mobility & requests– Not data‐popularities

Xenofon Vasilakos ‐ [email protected] 4

Page 5: Efficient Proactive Caching for Supporting Seamless Mobility · Efficient Proactive Caching (EPC) • Individual requests imply higher demand for cache space • Congestion pricing

Efficient Proactive Caching (EPC)

• Individual requests imply higher demand forcache space

• Congestion pricing for cache storage– Efficient cache utilization

EPC trades cache space (price) for reduced delay (delay cost)

Xenofon Vasilakos ‐ [email protected] 5

Page 6: Efficient Proactive Caching for Supporting Seamless Mobility · Efficient Proactive Caching (EPC) • Individual requests imply higher demand for cache space • Congestion pricing

Outline

1. EPC in a flat cache structure2. EPC in a two‐level cache hierarchy3. Evaluation

Xenofon Vasilakos ‐ [email protected] 6

Page 7: Efficient Proactive Caching for Supporting Seamless Mobility · Efficient Proactive Caching (EPC) • Individual requests imply higher demand for cache space • Congestion pricing

Flat cache structure

• Decision Rule:

Autonomous prefetching/ cachingXenofon Vasilakos ‐ [email protected] 7

Page 8: Efficient Proactive Caching for Supporting Seamless Mobility · Efficient Proactive Caching (EPC) • Individual requests imply higher demand for cache space • Congestion pricing

Flat cache structure

• Step‐wise decision procedure– Optimal selection of cached objects?

Xenofon Vasilakos ‐ [email protected] 8

Page 9: Efficient Proactive Caching for Supporting Seamless Mobility · Efficient Proactive Caching (EPC) • Individual requests imply higher demand for cache space • Congestion pricing

Flat cache structure

• Optimal selection of cached objects?1. Objects with different sizes

• Optimization is identical to 0/1 Knapsack Problem• NP‐hard problem

2. Optimal for equal‐size objects• For each cache and each request , order by 

Xenofon Vasilakos ‐ [email protected] 9

Page 10: Efficient Proactive Caching for Supporting Seamless Mobility · Efficient Proactive Caching (EPC) • Individual requests imply higher demand for cache space • Congestion pricing

Hierarchical cache structure

Xenofon Vasilakos ‐ [email protected] 10

Page 11: Efficient Proactive Caching for Supporting Seamless Mobility · Efficient Proactive Caching (EPC) • Individual requests imply higher demand for cache space • Congestion pricing

Hierarchical cache structure

• Leafs solve 2 flat cache problems :1. Delay DR

2. Delay DM

Xenofon Vasilakos ‐ [email protected] 11

Requires cooperation

Page 12: Efficient Proactive Caching for Supporting Seamless Mobility · Efficient Proactive Caching (EPC) • Individual requests imply higher demand for cache space • Congestion pricing

Finding an optimal solution ?

• Data  Placement Problem – Different object sizes => NP‐complete– Equal size objects => high polynomial degree time

Xenofon Vasilakos ‐ [email protected] 12

Page 13: Efficient Proactive Caching for Supporting Seamless Mobility · Efficient Proactive Caching (EPC) • Individual requests imply higher demand for cache space • Congestion pricing

EvaluationComparison with a naive, an optimal, and an oracle scheme

Xenofon Vasilakos ‐ [email protected] 13

Page 14: Efficient Proactive Caching for Supporting Seamless Mobility · Efficient Proactive Caching (EPC) • Individual requests imply higher demand for cache space • Congestion pricing

EvaluationComparison with a naive, an optimal, and an oracle scheme

Xenofon Vasilakos ‐ [email protected] 14

Page 15: Efficient Proactive Caching for Supporting Seamless Mobility · Efficient Proactive Caching (EPC) • Individual requests imply higher demand for cache space • Congestion pricing

EvaluationComparison with a naive, an optimal, and an oracle scheme

Xenofon Vasilakos ‐ [email protected] 15

Page 16: Efficient Proactive Caching for Supporting Seamless Mobility · Efficient Proactive Caching (EPC) • Individual requests imply higher demand for cache space • Congestion pricing

Evaluation

Xenofon Vasilakos ‐ [email protected] 16

Page 17: Efficient Proactive Caching for Supporting Seamless Mobility · Efficient Proactive Caching (EPC) • Individual requests imply higher demand for cache space • Congestion pricing

Evaluation

Xenofon Vasilakos ‐ [email protected] 17

Comparison with a naive, an optimal, and an oracle scheme

Page 18: Efficient Proactive Caching for Supporting Seamless Mobility · Efficient Proactive Caching (EPC) • Individual requests imply higher demand for cache space • Congestion pricing

A distributedmobility support solutiontailored to individual user mobility/requests that

exploits user mobility and uses congestion pricing

Xenofon Vasilakos ‐ [email protected] 18