 ICACT20260049 Slide.01
[Big Slide]
[YouTube] |
Chrome  Click!! |
 |
Hello and welcome to the presentation of my project, Creating A Scalable Smart Healthcare System Through The Utilization Of Edge/Fog Computing. My name is Rhaed Khiati and I am a PhD Candidate at TMU (Toronto Metropolitan University). I look forward to providing an innovation to the field with this project. |
| ICACT20260049 Slide.02
[Big Slide]
[YouTube] |
Chrome Click!! |
.jpg) |
The following is what I will be presenting to you today. I will introduce the problem with modern smart healthcare systems, while offering my potential solution and both the pros and cons of said solution. I will then go over the research expectations and expected contributions of this project in the long-term. I will then go over some of the literature available on this topic, before proceeding into discussing the prototype model I will be using, before concluding the presentation. |
| ICACT20260049 Slide.03
[Big Slide]
[YouTube] |
Chrome Click!! |
.jpg) |
Transition to Introduction. |
| ICACT20260049 Slide.04
[Big Slide]
[YouTube] |
Chrome Click!! |
.jpg) |
The production and advancement of our ever-growing technological landscape has resulted in the development of more modern smart healthcare systems and subsequently hospitals
Integration of AI and the Internet of Things (IoT) among others has resulted in a better quality of life for patients
Unfortunately, as these systems increase in scope and complexity, so does the number of patients using these systems
As a result, I am proposing a more sustainable model for these smart healthcare systems using a combination of edge and fog computing as the basis for this new model
|
| ICACT20260049 Slide.05
[Big Slide]
[YouTube] |
Chrome Click!! |
.jpg) |
Edge and fog computing both provide a number of benefits to smart healthcare systems, like an additional level of processing and filtering of patient data as it comes in from the various sensors and other IoT devices, reduced latency, enhanced fault-tolerance and so on
Ultimately, the addition of edge and fog computing does not make this project particularly ambitious or unique on its own
The application of Apache Beam, a distributed streaming framework, to the system as well to help alleviate a sizable number of the concerns that come with edge and fog computing
|
| ICACT20260049 Slide.06
[Big Slide]
[YouTube] |
Chrome Click!! |
.jpg) |
Transition to Problem Statement |
| ICACT20260049 Slide.07
[Big Slide]
[YouTube] |
Chrome Click!! |
.jpg) |
The proposed research essentially aims to address and answer the following research questions:
Can we design a scalable smart healthcare system using edge/fog computing and streaming frameworks?
Can the integration of Apache Beam further enhance the capabilities of edge/fog nodes in terms of big data management and scalability?
Could the creation of a dynamic scheduler by using ML techniques lead to a more efficient and less complex healthcare system?
|
| ICACT20260049 Slide.08
[Big Slide]
[YouTube] |
Chrome Click!! |
.jpg) |
The goal of this project is to create a more scalable and reliable smart healthcare system through the usage of some of the more prominent systems utilized in IoT and smart technology today
Hospitals today rely on vast amounts of big data, hence the need for a scalable solution to this ever-increasing concern
|
| ICACT20260049 Slide.09
[Big Slide]
[YouTube] |
Chrome Click!! |
.jpg) |
To achieve this goal, the following objectives must be met:
The characterization of a typical intelligent health care application, and obtaining an open source patient dataset for use in initial testing and simulation of the aforementioned application
The development and implementation of a custom scheduler that will be used to assign tasks and resources based on the tasks deadline
Connecting the scheduler to Apache Beam to manage the big streaming data
|
| ICACT20260049 Slide.10
[Big Slide]
[YouTube] |
Chrome Click!! |
.jpg) |
Transition to Expected Contribution |
| ICACT20260049 Slide.11
[Big Slide]
[YouTube] |
Chrome Click!! |
.jpg) |
This proposal aims to answer the first two research questions, by completing the following steps:
Building a prototype of the system, in which data is collected from various IoT devices, pooled together and sent to the various edge/fog nodes along the path towards the cloud
Apache Beam must be implemented prior to the data being sent to the edge/fog nodes as well as the cloud, and act as a relay hub broadcasting instructions for processing tasks to each node as well as analyzing the data steam and relaying information to the aforementioned nodes
|
| ICACT20260049 Slide.12
[Big Slide]
[YouTube] |
Chrome Click!! |
.jpg) |
This proposal aims to answer the first two research questions, by completing the following steps:
A scheduler would have to be implemented alongside Apache Beam to ensure that tasks are executed based on their impending deadline
Through characterization of the workload of the prototype, we can use simulation as the main method to measure the performance of the system
The third research question will potentially be implemented and answered once the first two have been completed and implemented
|
| ICACT20260049 Slide.13
[Big Slide]
[YouTube] |
Chrome Click!! |
.jpg) |
The following represents the expected contributions of this work to the field:
Creation of a system prototype to test and evaluate an intelligent health care system utilizing an edge/fog computing framework, a custom scheduler and Apache Beam
Creation of a scheduler and optimal task scheduling strategy that will be updated and enhanced over the course of development, that will be both ubiquitous and unique to this type of system
Adding machine learning and AI methods to enhance the validity and to provide a more comprehensive and complete smart system, specifically through the development of a unique and dynamic scheduler that utilizes said AI techniques
|
| ICACT20260049 Slide.14
[Big Slide]
[YouTube] |
Chrome Click!! |
.jpg) |
Transition to Literature Review |
| ICACT20260049 Slide.15
[Big Slide]
[YouTube] |
Chrome Click!! |
.jpg) |
The key point about this project is essentially the idea that this project is an amalgamation of several tried and tested concepts and systems that have been combined to form a singular and novel entity that has not been tried and tested to date
A number of the related works concern themselves with one or a couple of the ideas that are presented in this proposal, namely:
Fog/edge computing in smart cities
Usage of Apache Beam
Scheduling of tasks in smart healthcare systems
|
| ICACT20260049 Slide.16
[Big Slide]
[YouTube] |
Chrome Click!! |
.jpg) |
The first area that is a focal point of this proposal is the concept of fog/edge computing and its relation to smart cities
Based on the relevant research in the field, fog computing is utilized as it offers a number of benefits for smart healthcare, like:
Improved Mobility
Enhanced interoperability
Lower Latency
Better fault-tolerance
Improved network bandwidth
|
| ICACT20260049 Slide.17
[Big Slide]
[YouTube] |
Chrome Click!! |
.jpg) |
The second area that is a focal point of this proposal is the presence of Apache Beam
Beam is a streaming engine developed by an open source community under the umbrella of Apache, though initially developed by Google, one capable of real-time data processing of millions of events per second, using the least amount of resources. Benefits include:
Usage of a singular unified API for all cases (batch and streaming)
Ability to execute pipelines in a variety of runtime environments
Ability to write and create custom SDKs and transformation libraries
|
| ICACT20260049 Slide.18
[Big Slide]
[YouTube] |
Chrome Click!! |
.jpg) |
The third area that is a focal point of this proposal is the scheduling of tasks primarily in smart healthcare systems
Based on the relevant research in the field, schedulers offers a number of benefits for smart healthcare, like:
Better resource utilization and allocation
Enhanced time management in systems
Improved network bandwidth and less delays
Reduced the need for computationally complex systems
Improved load balancing
|
| ICACT20260049 Slide.19
[Big Slide]
[YouTube] |
Chrome Click!! |
.jpg) |
Transition to Methodology |
| ICACT20260049 Slide.20
[Big Slide]
[YouTube] |
Chrome Click!! |
.jpg) |
Initially, Apache Flink was considered in lieu of Beam, due to some of the known limitations that edge/fog nodes have with regards to the increased complexity on a system, and due to my personal familiarity with it
Flink offers a number of advantages like its use of event bookmarking to deal with fault-tolerance issues by rolling back to a specific ‘event’ determined by the parameters that were set, and the ability to handle data in real-time
|
| ICACT20260049 Slide.21
[Big Slide]
[YouTube] |
Chrome Click!! |
.jpg) |
Apache Flink does have its own glaring flaws, namely its individualistic nature, the necessity to run an independent copy in each node, and a lack of communication between APIs
This caused the formation of a bottleneck and made parallelization impossible
Results were also contradictory and often illogical
These were corrected through the use of Beam
|
| ICACT20260049 Slide.22
[Big Slide]
[YouTube] |
Chrome Click!! |
.jpg) |
Beam is portable and requires one API
Flink can still be used, in a ‘master-worker’ relationship
Data is collected from various sources, and sent to scheduler to pre-schedule tasks
Beam relays information once received
Data is pipelined and tasks are executed based on deadline
|
| ICACT20260049 Slide.23
[Big Slide]
[YouTube] |
Chrome Click!! |
.jpg) |
Figure 1: Prototype Model of the System |
| ICACT20260049 Slide.24
[Big Slide]
[YouTube] |
Chrome Click!! |
.jpg) |
Each edge and fog node receives their assigned tasks and data stream segment from the scheduling layer, which are received by the messaging queue
The queue then relays the information to the processor of the edge/fog node, which in this particular instance, is Apache Flink
Once analysis of the data and the submitted tasks are completed, Flink sends the data back to the messaging queue, which maintains constant communication with both Flink and the resource manager, who then relays this to any other edge or fog nodes
|
| ICACT20260049 Slide.25
[Big Slide]
[YouTube] |
Chrome Click!! |
.jpg) |
Figure 2: Internal look into each node |
| ICACT20260049 Slide.26
[Big Slide]
[YouTube] |
Chrome Click!! |
.jpg) |
With regards to the scheduler itself, for this proposal, a custom static scheduler is being implemented with the following considerations in mind:
Fog and edge nodes have different processing capabilities and loads, so a scheduler must take that into account when assigning or delegating tasks to each node
Depending on the priority level of certain tasks, they might have to be executed at different execution times or within a specific timeframe
Different resources must be allocated by the scheduler, which would also determine the node said task would be sent to
Nodes should be able to push/pull tasks based on circumstance
MIPS should also be taken into consideration
|
| ICACT20260049 Slide.27
[Big Slide]
[YouTube] |
Chrome Click!! |
.jpg) |
Transition to Conclusion |
| ICACT20260049 Slide.28
[Big Slide]
[YouTube] |
Chrome Click!! |
.jpg) |
Healthcare systems are and will continue to be a relevant topic of interest
The combination and adoption of fog computing, Apache Beam, and task scheduling into one unique and compact idea has the potential to bring significant change to the industry
In future developments, I plan to utilize features like machine learning/AI and a more dynamic scheduler to help simplify and automate the process, which is especially critical for the development of smart systems
By applying and completing the following goals, this project should be helpful to all manner of healthcare providers/stakeholders and potentially even save lives along the way
|
| ICACT20260049 Slide.29
[Big Slide]
[YouTube] |
Chrome Click!! |
.jpg) |
Thank you for your time.
If you have any questions, you may reach me at:
rkhiati@torontomu.ca
|
