使用地震传感器模拟一组互连的无线传感器网络（WSN），需要使用MPI来实现并发。
Preamble
A seismometer represents an instrument which responds to ground noises and shaking such as caused by earthquakes, volcanic eruptions, and explosions. These sensors are typically combined with a timing device and a recording device to form a seismograph. Seismometers play a crucial role in seismology, which is a scientific study of earthquakes. Figure 1 illustrates a screenshot or seismic activities which were recorded along Australia/New Zealand in August/September 2022. These activities provide vital information to governments on early warning of earthquakes. This link contains latest earthquakes in the Australia NZ Region as a reference on the type of data produced by these sensors (i.e., latitude, longitude, magnitude, depth, etc.)
However, the location or placement of seismometers are crucial as these sensors are sensitive towards noise and require careful calibration. As such, in an attempt to accurately detect underwater seismic activities, scientists have set up seafloor based seismic sensors.
Figure 2 illustrates a typical seafloor based seismic sensor setup, which consists of a surface buoy and a seafloor ocean bottom seismometer. Data collected from a seismometer could be transmitted to nearby seismometer or to a ground (or base) station for further processing.
In this assignment, you are required to simulate a wireless sensor network (WSN) of a set of interconnected seismic sensors. Figure 3 illustrates an example overview of the WSN and a balloon seismic sensor (aerial seismology). In this figure and for the purpose of illustration, nine seafloor based seismic sensors are positioned in a 3 3 Cartesian grid layout. The number of actual sensors may vary based on the area being observed. The sensors are positioned such that each sensor communicates with its immediate adjacent sensors (e.g., up, down, left, and right in the Cartesian grid layout). All seafloor seismic sensors in this grid can communicate with the base station and vice-versa. In Figure 3, the magenta-coloured dotted box/arrow shows that the seismic sensors send their readings to the base station (two-way communication).
In addition, a balloon floating in the Earth’s ionosphere observes the same Cartesian grid layout where the seafloor seismic sensors are currently positioned (as illustrated in Figure 3). This balloon uses a barometer to measure seismic readings, as part of a new method being researched to detect earthquakes using aerial seismology. The base station compares the received data from the seafloor seismic sensors and seismic readings from the balloon. This comparison is done to ascertain if there is a conclusive or an inconclusive match between the information sent by the seafloor seismic sensors and the balloon.
The aim of this group (of two or three members) assignment is to simulate the aforementioned setup. A simulation model needs to be designed and implemented using a combination of Message Passing Interface (MPI) and POSIX/OpenMP in C or C++ programming language. Detailed specifications are in the subsequent sections.
Objectives of the assignment:
- a) To design and implement a distributed event detection system based on simulating a wireless sensor network.
- b) To design and implement parallel algorithms in a distributed computing system.
- c) To analyse and evaluate the performance of the parallel algorithms and approaches in a distributed computing system.
- d) To apply technical writing to effectively communicate parallel computing.
Unit learning outcomes (LO) for this assignment:
- a) Design and develop parallel algorithms for various parallel computing architectures (LO3).
- b) Analyse and evaluate the performance of parallel algorithms (LO4).
- c) Apply technical writing and presentation to effectively communicate parallel computing to a range of academic and expert audiences (LO5).
WSN Description
The deployed wireless sensor network comprises m n nodes in a cartesian grid, and a base station. In the example illustration in Figure 2, m = 3 and n = 3. However, the values of m and n vary as specified by the user during program runtime. Each node in this cartesian grid simulates the behaviour of the seafloor seismic sensor, and each node communicates with its immediate adjacent nodes. These nodes can exchange data through unicast and broadcast modes of communications.
All nodes in the WSN can independently exchange data with the base-station. Base-station for this WSN is an additional computer node that is entrusted with the task of gathering data from all sensor nodes in the WSN. All communications among nodes, and between the nodes and the base station are achieved using Message Passing Interface (MPI). In essence, each sensor node and base station can be represented as a MPI process.
Simulating the seafloor seismic sensor node
- a) A single MPI process simulates one sensor node. For instance, in a 3 3 Cartesian grid layout, there will be 9 MPI processes. Each node will have a (x, y) coordinate (e.g., Node 1 - (0,0), Node 2 - (0,1), Node 3 - (0,2), Node 9 (2, 2)). You could use MPI virtual topologies here.
- b) Each node periodically produces a simulated sensor reading. This reading consists of date, time, latitude and longitude of the earthquake point, magnitude, and depth of the quake from the sensor.
You can generate random float values to simulate the sensor readings and you can refer to this link as a reference. The following table lists sample readings (simulated at a cycle of 10 seconds): - c) If the generated magnitude at each time interval average exceeds a predefined threshold, this constitutes a possible earthquake. The node will then send a request to its immediate adjacent neighbourhood nodes to acquire their readings for comparison purposes. To reiterate, the neighbourhood nodes refer to immediate top, bottom, right and left adjacent nodes.
- d) Upon receiving the readings from its neighbourhood nodes, the node compares these readings to its own reading to check if the readings are similar. Comparisons here are based on:
- i. Computing the absolute difference in latitude and longitude coordinates of reported seismic events between the nodes.
- ii. Computing the absolute difference in magnitude readings of reported seismic events between the nodes.
- e) Should the difference in location and magnitude from at least two or more neighbourhood nodes match the readings of the local node (within a threshold which you can determine), the node sends a report (i.e., alert) to the base station. To compute the distance between two geographical coordinate points in C, you may refer and use the code in this link with proper citation to the code in your report.
- f) The report sent to the base station should contain as much information as possible about the possible alert. This information includes the timestamp (including the date & time) at which an alert is detected, sensor value readings (e.g., magnitude, depth, location) and number of messages compared with the neighbourhood nodes. You should demonstrate efficiency when reporting an alert message to the base station. In this context, you should minimize the number of calls to the MPI Send or Isend functions by a node to the base station when reporting an alert condition.
- g) Each node repeats parts (a) to (g) until upon receiving a termination message from the base station. Once the node receives a termination message, the node cleans up and exits.
- h) Simulation should work for dynamic value of m x n nodes and threshold settings. At start up, the program allows the user to specify the grid size (m x n) and threshold values.
- i) If you are aiming for HD or upper HD: The node uses a thread (i.e., POSIX or OPENMP) to send or receive MPI messages between its adjacent nodes. This thread is created by the sensor node and terminates properly at the end of the program.
Simulating the balloon seismic sensor
- a) A single POSIX thread simulates the balloon seismic sensor (you may opt to use OpenMP as an alternative to POSIX thread). Note that the balloon covers the same area of the seafloor seismic sensors. For simulation purposes, this thread is created by the base station node (refer to point 3.0 below for details about the base station).
- b) This thread periodically produces seismic readings. This reading consists of date, time, latitude and longitude of the earthquake point, magnitude, and depth of the quake from the sensor. However, the generated magnitude always exceeds the predefined threshold
- c) The following table lists sample seismic readings produced by the thread (simulated at a cycle of 1 second) for the entire Cartesian grid layout.
- d) The information in (c) is stored in a shared global array, which can also be accessed by the base station node. The array has a fixed size, and you can decide the size of this array. Once the array is full, the thread removes the first entered data from the array to make way for the new data (first in, first out approach).
- e) The thread repeats (a) to (d) until upon receiving a termination signal from the base station. Once the thread receives a termination signal, the thread exits.
Simulating the base station
- a) A single MPI process simulates the base station node.
- b) The base station node also creates the thread which simulates the balloon seismic sensor (refer to point 2.0 above).
- c) The base station node periodically listens for incoming reports from the seafloor seismic sensor nodes.
- d) Upon receiving a report from a sensor node, the base station compares the received report with the data in the shared global array (which is populated by the balloon seismic sensor).
- If there is a match, the base station logs the report as a conclusive event (conclusive alert).
- If there is no match, the base station logs the report as an inconclusive event (inconclusive alert).
- Note: You can set your own threshold values when comparing the content of the received report and data in the shared global array. However, comparisons between the node report and shared global array should factor in the difference in location and magnitude. You could also include depth as a third comparison factor.
- e) The base station writes (or logs) the key performance metrics (e.g., the simulation time, number of alerts detected, number of messages/events with senders’ adjacency information/addresses, total number of messages (for this simulation)) to an output file. Doing so may assist in proving the correctness and efficiency of your implementation. You can also include IP addresses of the node and/or base stations, but including a loopback address (127.0.0.1) is not considered here.
- f) The base station program allows the user to specify a sentinel value at runtime to allow a proper shutdown of the base station, balloon sensor and sensor nodes. Note: CTRL+C is not allowed.
- g) Continuing from (f), the base station sends a termination message to the sensor nodes to properly shutdown. The base station also sends a termination signal to the balloon seismic sensor to properly shutdown.
- h) If you are aiming for HD or upper HD: The base station uses a thread (i.e., POSIX or OPENMP) to send or receive MPI messages from the sensor nodes. This thread is created by the base station and terminates properly at the end of the program.
Fault detection (Only for teams of three members)
- a) A fault detection mechanism is simulated whereby the sudden failure of a node is detected by the base station and/or adjacent sensor nodes.
- b) Once a faulty node is detected, a method is implemented to notify the base station to log the fault and to gracefully shutdown the entire network.
Note: Fault detection is for teams of three members. If you belong to a team of two members, you are not required to implement this feature.
Important: The aforementioned description represents the baseline in meeting the requirements of the assignment. The marking rubric will include specific requirements to obtain a HD for a criteria in this assignment.
Submission
Teamwork: Yes. You will work in a team of two (or three) members to complete this assignment. Your team formation is based on the same team formation used in the weekly lab activities and in the first assignment. Each team member within a group will produce the same set of report, code files, log files and any other supplementary materials. However, each team member is required to make a submission independently in Moodle.
Programming Task: Design, write and demonstrate an MPI + thread program that effectively addresses all the points in Section II (WSN Description). Assume that a set of MPI processes represents the WSN and each MPI process represents a WSN node and the base station.
Write-up Task - E-folio style report layout:
- Methodology - Description of the methods and algorithms used to simulate the seafloor seismic sensor nodes, balloon seismic sensor and base station. Include illustration to describe the design, and flowchart (or pseudocode) of the algorithm(s).
- Results tabulation - Tabulation or illustration of the results using tables, graphs, or screenshots of the log file content.
- Analysis & discussion - Provide an analysis on the behaviour of the simulator. Describe the content of the messages which were exchanged between the nodes and messages that were sent to the base. Include any additional observations to support the analysis.
Maximum number of words in the report: 1,500 words for two member teams, 2,000 words for three member teams. The word count excludes figures, tables, and references (if any).
- Demo Interview Task: The group needs to attend a demo and interview session held during Week 12 lab session. The interview is individually assessed.
Please refer to FAQ which are enclosed with the assignment specifications in Moodle. In addition, please refer to the assessment rubric which is enclosed together with this specification. This rubric should assist you to attain a better understanding of the marking guide for the assignment.