Multi-User Computation Partitioning for Latency Sensitive Mobile Cloud Applications

Question

1. What are the contributions mentioned in the paper "Multi-user computation partitioning for latency sensitive mobile cloud applications" ?

2. What is the objective of the optimization in MCPP?

3. What is the problem to partition computations on a constrained number of cloud resources?

4. How many occupied servers are there at this interval?

Accepted Answer

Existing works focus on the single-user computation partitioning, which aims to optimize the application completion time for one particular single user.. In this paper, the authors study, for the first time, Multi-user Computation Partitioning Problem ( MCPP ), which considers the partitioning of multiple users ’ computations together with the scheduling of offloaded computations on the cloud resources.. Instead of pursuing the minimum application completion time for every single user, the authors aim to achieve minimum average completion time for all the users, based on the number of provisioned resources on the cloud.. The authors show that MCPP is different from and more difficult than the classical job scheduling problems.. The authors demonstrate through benchmarks that SearchAdjust outperforms both the single user partitioning approaches and classical job scheduling approaches by 10 % on average in terms of application delay.. Based on SearchAdjust, the authors also design an online algorithm for MCPP that can be easily deployed in practical systems.

Accepted Answer

In TSPHC the optimization objective is the makespan which is the maximum completion time of all the tasks, while in MCPP the objective is the total weighted completion time.

Accepted Answer

The problem is to schedule the offloaded computations on a constrained number of cloud resources as well as to partition the computations between mobile side and cloud side for all the users, such that the average application delay is minimized.

Accepted Answer

In the algorithm, Lcro is first initialized by the time interval (0,∞), with the number of occupied servers at this interval being zero.

Accepted Answer

jc represents the first successor of module j0 that is scheduled to the cloud; if no successor of module j is at the cloud side, then τ ′i,jc = ∞. jm is the first successor of module j0 that is scheduled onto the mobile side; if no successor of module j0 is at mobile side, then jm = n+ 1.Extra Delay.

Accepted Answer

The reason is that as var increases, the load fluctuation over time has greater magnitude, resulting more time slots in which the cloud servers are either overloaded or mostly spare.

Accepted Answer

SearchAdjust was demonstrated to outperform the list scheduling algorithms, HEFT and MEDLS, by 10 percent in term of the application delay.

Accepted Answer

For the MCPP, with the optimal partitions xi,j,0 for each user i, the authors can easily obtain the initial execution schedule S = {(xi,j,0, τi,j , ti,j)}.

Accepted Answer

This evaluation implies that their online algorithm can achieve a delay of 1.08dopt at most time, i.e. 90% time of the whole month in October 2007 for wikipedia requests data set.

Accepted Answer

Through the evaluation of various offline solutions, the authors aim to answer two questions: 1) which solution performs best; 2) how the application performance (delay) varies depending on the number of cloud resources and the load, i.e., PerformanceResource-Load (PRL) Model.

Accepted Answer

This is because the greedy heuristics need more iterations to adjust the SCPP initial solution as the cloud resources constraint is lower.

Accepted Answer

For the s-th module which comes to the cloud in s-earliest time, the authors can prove if 1-st, 2-nd, ... , (s − 1)-th modules start to run at their arriving time, then the s-th module can also be executed at the cloud servers at its arriving time τis,js .

Accepted Answer

One way is, as described in 3.4, the authors can fit the MCPP into the TSPHC model by abstracting both the mobile devices and cloud servers as the processors.

Accepted Answer

The authors can prove using mathematical induction, by the ’first-come-first-serve’ policy every offloaded module (i, j) can start exactly at time τi,j and be completed by the time ti,j on the cloud servers.

Accepted Answer

Another interesting thing the authors can observe from Fig.5c is that, their proposed greedy heuristics have less running time as the cloud resources increase, while HEFT has longer running time as the cloud resources increase.

Multi-User Computation Partitioning for Latency Sensitive Mobile Cloud Applications

Chat with Paper

AI Agents for this Paper

Most frequently asked questions

1. What are the contributions mentioned in the paper "Multi-user computation partitioning for latency sensitive mobile cloud applications" ?

2. What is the objective of the optimization in MCPP?

3. What is the problem to partition computations on a constrained number of cloud resources?

4. How many occupied servers are there at this interval?

5. what is the first successor of module j0 that is scheduled to the cloud?

6. Why is the application delay of the wikipedia request data set more than a var?

7. How much does SearchAdjust outperform the list scheduling algorithms?

8. How do the authors get the initial execution schedule for the MCPP?

9. How long does the wikipedia request delay take to be able to be accessed?

10. What is the purpose of the evaluation of offline solutions?

11. Why is the greedy heuristic better than MEDLS?

12. how can i prove that a module is executed at the cloud?

13. How can the authors fit the MCPP into the TSPHC model?

14. how can the authors prove that the offloaded modules are completed on the cloud?

15. What is the difference between the two greedy heuristics?

Figures

Citations

A Survey on Mobile Edge Computing: The Communication Perspective

A Survey on Mobile Edge Computing: The Communication Perspective

Mobile-Edge Computing: Partial Computation Offloading Using Dynamic Voltage Scaling

Joint Task Offloading and Resource Allocation for Multi-Server Mobile-Edge Computing Networks

Offloading in Mobile Edge Computing: Task Allocation and Computational Frequency Scaling

References

Distinctive Image Features from Scale-Invariant Keypoints

Scheduling: Theory, Algorithms, and Systems

The Case for VM-Based Cloudlets in Mobile Computing

Performance-effective and low-complexity task scheduling for heterogeneous computing

MAUI: making smartphones last longer with code offload

Related Papers (5)

Efficient Multi-User Computation Offloading for Mobile-Edge Cloud Computing

The Case for VM-Based Cloudlets in Mobile Computing

A Survey on Mobile Edge Computing: The Communication Perspective

Energy-Efficient Resource Allocation for Mobile-Edge Computation Offloading

Mobile Edge Computing: A Survey on Architecture and Computation Offloading