08341c3681
Minor Modifications: * Indentation issues are fixed * Typo errors in source code and comments are fixed * Misspelled parameters in plotTaskFailureReason.m are corrected * sim_results folder is added to .gitignore file Backward compatible changes * The exit code of the application when there is an error is changed from 0 to 1 * Default constructors are added for Location.java, LoadGeneratorModel.java, MobilityModel.java, EdgeOrchestrator.java (this change request coming from a developer) * public TaskProperty(int mobileDeviceId, double startTime, ExponentialDistribution[] expRngList) is added to TaskProperty.java (this change request coming from a developer to create a task without task type) * double getCreationTime() function is added to Task.java * void reconfigureMips(double mips) function is added to EdgeVM (for future usage) * gsm_propagation_delay variable is added to config file. SimSettings class is also modified accordingly. You can use it if you have cellular network access in your scenario. * wlan_range, northern_bound, southern_bound, eastern_bound, western_bound variables are added to config file; and relevant functions are added to SimSettings class. (this change request coming from a developer) Backward incompatible changes! * location_check_interval variable name is changed to location_check_interval in config.properties file. Please update your config files accordingly (remove 'vm_' part) * Major modifications are applied in SimLogger class to decrease time complexity. Now the basic results are kept in the memory and saved to the files at the end of the simulation. As a result of this change, the signature of the SimLogger.addLog () function had to be changed. You must add the mobile device id as the first argument. Please update your MobileDeviceManager class accordingly (add task.getCloudletId () as the first argument).
Sample Application 4
This application includes the source code which is used in our paper submitted to IEEE Transactions on Network and Service Management [1].
The digital object identifier of the paper will be provide after it is published!
Fuzzy Logic Based Workload Orchestrator
In this application we introduce a fuzzy logic based workload orchestrator. In our design, two stage fuzzy logic system is used as shown in Figure 1. In the first stage, the most convenient edge server in the edge layer is found. In the second stage, the candidate edge server and the cloud server are compared. As a result of these operations, the proposed fuzzy logic based workload orchestrator finds a target server which can be the local edge server, remote edge server or the cloud server.
Figure 1: Two stage fuzzy workload orchestration.
Why Fuzzy Logic
- Computation offloading is NP-Hard with many inputs and output variables
- Fuzzy logic easily handles multi-constraint optimization process
- There is no simple decision model for offloading
- Fuzzy logic lets us handle the uncertainty with simple if-then rules
- Fuzzy logic abstracts complex operational details in a high level human understandable format
Simulated Environment
A university campus environment is simulated in this application. The students demand services from the edge servers within the campus while moving around the predesignated locations such as the classroom buildings, cafeteria, dormitories, and administrative offices. The individuals are assumed to carry and/or wear the mobile gadgets which continuously run applications that in turn create a traffic due to offloading of the tasks. There exists certain number of edge servers located in the campus such that they provide coverage for edge computing requests. A user is connected to the nearest edge server via WLAN whereas the edge servers are interconnected by MAN. Also, the standard cloud infrastructure can be accessed over WAN. In this setting, it is possible for the user to offload a task to (i) Nearest edge server (we call this the local edge server) (ii) a neighboring edge server within the campus connected by AN (we call this the remote edge server), and (iii) to the global cloud servers. In our work, this decision is handled by the workload orchestrator.
Figure 2: A university campus scenario representing multiple locations.
References
[1] IEEE Transactions on Network and Service Management, [Online]. Available: https://www.comsoc.org/publications/journals/ieee-tnsm. [Accessed: 21- Jan- 2019].