ICACON Mobile Application Offloading: An Opportunity towards Mobile Cloud Computing. A. Ellouze, M. Gagnaire. May 22, 2015

Transcription

1 ICACON 2015 Mobile Application Offloading: An Opportunity towards Mobile Cloud Computing A. Ellouze, M. Gagnaire May 22, 2015

2 Outline Research Motivation Offloading decision model Decomposition of energy consumption State of charge of the battery Mobile Application Offloading Algorithm (MAO) Numerical Simulation Future Work A. Ellouze, M. Gagnaire ICACON 2015 Institut Mines-Télécom 2 / 16

3 Research Motivation Motivation An offloading strategy for mobile applications; Infrastructure deployment cross optimization tool between Radio resources disponibility and VMs allocation; A. Ellouze, M. Gagnaire ICACON 2015 Institut Mines-Télécom 3 / 16

4 Offloading decision model Mobile Processor Job Modeling Figure: Processor Sharing Model A. Ellouze, M. Gagnaire ICACON 2015 Institut Mines-Télécom 4 / 16

5 Offloading decision model Offloading Decision Model Concept of critical delay: Expected excution delay D expected (j) given N(j),t a(j), C M Ideal execution delay D(j) Preferred waiting delay (in relation with QoE) d(j) Critical delay D (j) = D(j) + d(j) A new application is eventually offloaded if: D expected (j) > D (j) (1) A. Ellouze, M. Gagnaire ICACON 2015 Institut Mines-Télécom 5 / 16

6 Decomposition of energy consumption Decomposition of energy consumption If job j is processed on the mobile terminal, the minimum energy consumption E M min(j) is given by: E M min(j) = P active M D M min(j) (2) If this same job is processed on the server, the energy E S (j) consumed by the mobile terminal for this offload is given by: E S (j) = E tr (j) + E M idle,p(j) + E M idle,w(j) (3) where, E tr (j) = P V (j) B P(P tr ); P tr = min(p max, P log 10 (M) + PL) (5) (4) Eidle,p(j) M = N(j) PM idle (6) C S Eidle,w(j) M = ρ N(j) C S 2 (1 ρ) Pidle M (7) A. Ellouze, M. Gagnaire ICACON 2015 Institut Mines-Télécom 6 / 16

7 State of charge of the battery State of charge of the battery 20% as a threshold to offload applications meeting the requirements. SoC = 1 αa, where, α α f A is the accumulated capacity during the time period [t s, t e] at the discharge current rate I and α f is the full capacity. Figure: Battery model 1 1 J. C. Zhang, S. Ci and H. Sharif, "An Enhanced Circuit-Based Model for Single-CellBattery," IEEE Applied Power Electronic Conference and Exposition (APEC), February 21-25, 2010, Palm Springs, pp A. Ellouze, M. Gagnaire ICACON 2015 Institut Mines-Télécom 7 / 16

8 Mobile Application Offloading Algorithm (MAO) Algorithm Flowchart Main contributions: QoE Satisfaction; Network conditions considered; User Position within the Cell considered; CPU considered; State of Charge of the battery considered; Figure: MAO Flowchart. A. Ellouze, M. Gagnaire ICACON 2015 Institut Mines-Télécom 8 / 16

9 Mobile Application Offloading Algorithm (MAO) Applications - Benchmark Range of applications: Chess game: processing inherent to the execution of one move on the chess-board; Speech recognition: conversion of an analog speech to a text and vice versa; Virus scanning: a program able to detect the presence of virus, and if possible, to kill it; Table: Applications parameters. Application CPU Interactivity d(j) in D(j) in seconds seconds Chess game High High Speech recognition Low High Virus Scanning High Low A. Ellouze, M. Gagnaire ICACON 2015 Institut Mines-Télécom 9 / 16

10 Numerical Simulation Simulation parameters Table: Simulation parameters. 2 Parameter Value Cellular layout 1 Single hexagonal cell Cell radius 500 m Path loss model PL = A+30 log 10 ( d d0 f +X h +σ Channel fading Typical urban Carrier frequency 2GHz System Bandwidth 5MHz (25RBs) Thermal Noise per RB dbm Bit rate of the mobile 600Kbps < B < 16Mbps Traffic Control messages, data traffic Number of UE 1 Number of application activation 100 Inter-arrival in seconds λ 1 2, 15, 30 Simulation duration 3 hours server. 2 we assume that an application server is located at the antenna s site and the three applications softwares are installed on the remote A. Ellouze, M. Gagnaire ICACON 2015 Institut Mines-Télécom 10 / 16

11 Numerical Simulation Equally mixed applications Variable loads Variable CPU Speed Energy Savings and Rejection Ratio under variable loads (Scenario Mix) Energy Gain Rejection Ratio Performance Evaluation With Different Computational Resources Energy Gain Rejection Ratio EnergyGain & RejectionRatio EnergyGain & RejectionRatio inter arrival of applications (in seconds) F (CPU Server Speed per CPU Mobile Speed) X axis: Inter-arrival in seconds Y axis: Energy Gain and Rejection Ratio X axis: F (Computational ratio) Y axis: Energy Gain and Rejection Ratio Conclusion: When inter-arrivals gets greater, offloading applications is less beneficial. Having more computational resouces provides avenues for more gains to be achieved in terms of energy savings. A. Ellouze, M. Gagnaire ICACON 2015 Institut Mines-Télécom 11 / 16

12 Summary Summary Developed Offloading decision model; Designed offloading algorithm tweaking on energy efficiency and QoE(MAO); Used realistic applications to simulate; Evaluated model through simulated results; A. Ellouze, M. Gagnaire ICACON 2015 Institut Mines-Télécom 12 / 16

13 Perspectives Future Work Recovery effect of the battery to take into account; Overhead virtualization and architectures s compatiblity to include; Cross optimzation tool between radio resources and VM s placement; A. Ellouze, M. Gagnaire ICACON 2015 Institut Mines-Télécom 13 / 16

14 Bibliography Bibliography Dusza B, Ide C, Cheng L, Wietfeld C. " An accurate measurement-based power consumption model for LTE uplink transmissions ", In Proc. IEEE INFOCOM (Poster), Turin, Italy, X. Ma, Y. Cui, L. Wang and I. Stojmenovic "Energy Optimization for Mobile Terminals via Computation Offloading", nd IEEE International Conference on Parallel Distributed and Grid Computing (PDGC), 2012, pp J. C. Zhang, S. Ci and H. Sharif, "An Enhanced Circuit-Based Model for Single-CellBattery," IEEE Applied Power Electronic Conference and Exposition (APEC), February 21-25, 2010, Palm Springs, pp Dusza, B., Ide, C., Cheng, L. and Wietfeld, C. (2013), CoPoMo: a context-aware power consumption model for LTE user equipment. Trans Emerging Tel Tech, 24: UE radio transmission and reception, January GPP TS , V E. Lagerspetz and S. Tarkoma, "Mobile search and the cloud: The benefits of offloading," in Ninth Annual IEEE International Conference on Pervasive Computing and Communications, PerCom 2011, March 2011, Seattle, WA, USA, Workshop Proceedings. IEEE, 2011, pp John L. Henning, SPEC CPU2006 benchmark descriptions, SIGARCH Comput. Archit. News34,September Architecture for LTE Mobile Terminals", IEEE Conference, A. Ellouze, M. Gagnaire ICACON 2015 Institut Mines-Télécom 14 / 16

15 ICACON 2015 Mobile Application Offloading: An Opportunity towards Mobile Cloud Computing A. Ellouze, M. Gagnaire May 22, 2015

16 The End Related Work Related Work E. Cuervo, A. Balasubramanian, D.-k. Cho, A. Wolman, S. Saroiu, R. Chandra, and P. Bahl, "Maui: making smartphones last longer with code offload," in Proceedings of the 8th international conference on Mobile systems, applications, and services" B.-G. Cn, S. Ihm, P. Maniatis, M. Naik, and A. Patti, "Clonecloud: elastic execution between mobile device and cloud." in EuroSys, C. M. Kirsch and G. Heiser, Eds. ACM, 2011, pp Shiraz, M. Gani, A. Khokhar, R.H. Buyya, R., "A Review on Distributed Application Processing Frameworks in Smart Mobile Devices for Mobile Cloud Computing," Communications Surveys and Tutorials, IEEE, vol.15, no.3, pp.1294,1313, Third Quarter 2013 A. Ellouze, M. Gagnaire ICACON 2015 Institut Mines-Télécom 16 / 16