Healthcare is a fundamental part of every individual’s life. The healthcare industry is developing very rapidly with the help of advanced technologies. Many researchers are trying to build cloud-based healthcare applications that can be accessed by healthcare professionals from their premises, as well as by patients from their mobile devices through communication interfaces. These systems promote reliable and remote interactions between patients and healthcare professionals. However, there are several limitations to these innovative cloud computing-based systems, namely network availability, latency, battery life and resource availability. We propose a hybrid mobile cloud computing (HMCC) architecture to address these challenges. Furthermore, we also evaluate the performance of heuristic and dynamic machine learning based task scheduling and load balancing algorithms on our proposed architecture. We compare them, to identify the strengths and weaknesses of each algorithm; and provide their comparative results, to show latency and energy consumption performance. Challenging issues for cloud-based healthcare systems are discussed in detail.
Healthcare is a fundamental part of every individual’s life. The healthcare industry is developing very rapidly with the help of advanced technologies. To ensure healthcare systems are more accessible to people, many researchers are trying to build different healthcare solutions. Nowadays, smartphones, as IoT mobile devices, are more capable of dealing with diverse types of applications to complete their tasks. Healthcare applications on mobile devices can exchange data through communication interfaces (e.g., application programming interfaces (APIs)) between patients/system users and healthcare service providers. However, the limitations of computing resources (e.g., CPU, storage, and processing power) in mobile devices mean that is not possible to run all application processes in these resource-constrained mobile devices. Thus, to overcome resource limitations, mobile devices integrated with cloud paradigms as a mobile cloud computing architecture offer efficiency-enhancing usability of the mobile device. For example, users can access healthcare facilities and get the task outputs by offloading computation activities from mobile devices with their hardware limitations to cloud-based techniques. Therefore, in this paper, we propose a hybrid mobile cloud computing (HMCC) architecture for healthcare applications. The HMCC architecture contains one or more private clouds where patient’s information can be stored and analyzed, and one or more public clouds for easy access to the healthcare system for patients. HMCC provides a workload balancing algorithm for the proper utilization of resources.
With the increase in population, healthcare systems are becoming major challenges in today’s world. According to the World Health Organization (WHO) [
These healthcare problems can be alleviated through the use of cloud computing techniques. Cloud computing has been widely revolutionized by incorporating computing technologies. Thus, using cloud computing provides the main benefits of: (1) enhancing the usability of existing IoT resources, (2) allowing users to access hardware components, such as storage or CPU, as well as software components, at any time from any location, (3) providing high-capacity networks, as well as low-cost computing and storage services, and (4) guaranteeing high-accuracy results, as well as requiring less human interaction.
According to cloud deployment models [
Based on the National Institute of Standards and Technology (NIST) cloud computing reference architecture [
However, cloud computing resources based on the cloud computing architecture in
The remainder of this paper is organized as follows. Section 2 provides a detailed literature survey of existing healthcare systems with cloud computing approach, highlighting the need for mobile cloud computing in health care systems. Section 3 details the proposed architecture and load balancing algorithms; then Section 4 discusses the implementation environment and analyzes the results. Based on our implementation and literature survey, Section 5 highlights open challenges; Section 6 then concludes the paper.
Health care services incorporate recent advancements in technology. Various applications leveraging remote photoplethysmography techniques, such as the remote analysis of patients using video and web cameras, are widely used. 3D remote computed tomography (CT) is another remote technique used nowadays for imaging and health care automation [
Reference [
Reference [
Collaboration among various healthcare systems remains an issue. Various pharmacies, hospitals, clinics, emergency services, and insurance companies all follow different naming systems. Reference [
An IoT-based healthcare application using PaaS prototype was proposed in Ref. [
Mobile cloud computing (MCC) is built based on concepts of cloud computing and mobile computing as the combination of cloud computing technologies with mobile devices, to bring rich computational resources to mobile users [
However, as is well know the size of a mobile device is usually small, the maximum capacity of computation, storage, and power is always limited. Thus, high-volume data processing in healthcare systems based on MCC is managed efficiently and synchronized into a distributed execution of cloud computation and mobile device. This MCC-based solution is a computation offloading, such that the data processing part would be sent to the cloud servers to be integrated, and then once the data execution tasks have been completed, sent back to the mobile device. From our earlier research [
Energy efficient scheduling on federated Edge cloud (ESFEC) in Ref. [
Among multiple issues highlighted for edge computing in healthcare systems, we propose an architecture to address the following issues:
Our proposed hybrid cloud has at least one private cloud, and at least one public cloud. The internal structures of the two types of clouds are consistent with each other. In our proposed architecture, private clouds are used to store and process medical data within the health organization. This allows Information Technology (IT) staff to have more control over stored medical data. Only physicians and IT staff have direct access to private clouds. Users can get their health data and diagnosis updated from that private cloud, but they need to be authorized members of that system. The public cloud is open to both physicians and patients. The patient’s and doctor’s simplified profiles are there. Our main objective is to maintain connectivity while the user is moving from one place to another place, and incorporating an efficient load balancing algorithm for the maximum usage of the cloud resources. We describe the scenario of this model in The mobile user will be connected to the main cloud through edge-clouds using mobile data, or an access point (Wi-Fi, hotspot, etc.) when the region is covered by an existing edge-cloud (or if connectivity is available, directly to the main cloud). When the user is at the edge of an edge cloud, and about to move out from that edge, they are connected to the nearest available edge cloud. The system migrates the ongoing process from the previously connected edge cloud to a newly connected edge cloud through the main cloud with minimum interruption. If the number of service requests in a specific edge cloud is more than its threshold level, the system automatically connects the upcoming services to the nearest available edge-cloud by dynamic load-balancing algorithms. If no edge-cloud is available during the mobility of user and after crossing the threshold of a particular edge-cloud, the system automatically creates a new edge-cloud. When the last user leaves the edge cloud, the system automatically drops the unused edge cloud. The load balancing algorithm produces minimum data overhead, and a management system measures and controls the data overhead of the system.
After the task offloading decision is made, the tasks from users are divided into two major categories; a set of tasks is offloaded to remote servers (edge and/or cloud), while other sets are computed locally. For those who are offloaded, they need one more level of optimization, which will help to further reduce the latency and energy consumption. This level is load balancing. Assume there are
Load balancing algorithms are normally two categories: static and dynamic. Static algorithms are suitable for low traffic data and this traffic is equally distributed over all servers. But when any server gets overloaded, migration does not depend on the current state of the system. Dynamic algorithms consider the current state of the system and distribute workload based on that. We studied static algorithms like round-robin, weighted round-robin, min-min, and max-min static algorithms [
Load balancing algorithm | Type | Benefits | Limitations |
---|---|---|---|
Energy efficient scheduling on federated edge cloud based on energy first (ESFEC-EF) [ |
Static |
This algorithm has lower energy consumption in high traffic networks. Despite its limitations, service migration overhead is reduced due to VM selection based on minimum overhead and maximum CPU utilization. |
The number of service migrations increases with the increase in traffic. |
Deep deterministic policy gradient (DDPG) [ |
Dynamic |
It can handle a more dynamic environment and scales well with substantial number of request/servers. It optimizes resource placement and task dispatching process lowering the overall latency and energy consumption. |
It takes more iterations and hence more running time to optimize the balancing problem. But the results are much better than static algorithms. Equilibrium should be attained among number of iterations and optimization objective value. |
Graph coloring (GRAPH) [ |
Dynamic |
It provides scalability to edge servers and cloud servers by increasing the CPU utilization. It helps with load balancing to reduce the overall latency of the system and reduce the network traffic. This method is good for dense traffic networks. |
Unlike latency, energy consumption for higher task sizes is more. This is due to more transmission and migration costs incurred to lower the latency. |
After observing the benefits and limitations of three load balancing algorithms in
The total time
The total energy
We compare three load balancing algorithms using our proposed architecture. For our proposed a hybrid mobile cloud computing (HMCC) architecture, load balancing for multiple tasks that are offloaded to edge servers is evaluated using an energy efficient scheduling on federated edge cloud based on energy first (ESFEC-EF) algorithm from Ref. [
We simulate the working for each algorithm using Python on 11 Gen Intel® Core™ i7-11370H @ 3.30 GHz, 2,995 MHz, 4 cores, and 8 logical processors.
Parameter | Description | Initialization |
---|---|---|
M | Maximum number of edge servers available | 20 |
N | Number of user devices scheduling task | [40, 90, 140, …, 390] |
B | Size of each task | Random (2 − 200) MB |
Computation cycles to process one byte on CPU | 1 | |
CPU frequency | 5 GHz | |
Energy consumed by CPU when idle | 75 W |
For increasing number of users with minimal task size (2 MB), we compare the latency consumed shown in
Development of web applications and mobile applications for remote access to healthcare professionals. These applications are intended for chronic diseases that need immediate diagnosis and treatment, like congestive heart failure, arrhythmia, and hypoxia. They use human–computer interface designing, text messaging, SMS, calls, text document sharing, etc. This approach lacks real-time collection of physiological signals and readings from patients. It does not support video calling features, or wireless communication. These features limit the scope of healthcare systems in wireless form. The number of patients reviewed per day remains less. It requires patients to have skills to collect and input data into applications. Thus, it remains useless in the case of emergency. This approach uses cloud servers for high traffic healthcare applications. Improvement of network level resources promotes availability and computation heavy healthcare applications to be used. This approach resolves several serious issues concerning security, data protection and ownership, quality of services, and mobility. This leads to expansion of capabilities and benefits and the overcoming of limitations, such as limited memory and CPU power. This approach requires mobile devices to communicate with cloud servers directly. Though the computation cost is reduced due to task offloading, the cost of transmission increases. This leads to latency, as well as energy consumption. With the increase in IoT devices in healthcare systems, there is a need for low energy consuming solutions. Edge computing offers useful computing resources at the edge of the network to maintain low-latency and real-time computing. Computing solutions are provided at the edge of the network. All patient data is stored on the same, which is accessed by healthcare professionals and patients from public network. This introduces security concerns. With readily available IoT devices, increased accessibility to mobile networks, network traffic increases, and edge servers are often overloaded. We separate the data stores and access rights for health professionals and public network, which deals with the concern of security. The major focus of our paper also lies in dynamic load balancing algorithms to optimize latency and energy consumption. This solution is suitable for low battery IoT devices, and mobile devices. Virtual edge cloud creation also helps when the capacity of available edge servers is all used up. Hybrid edge and cloud computing and load balancing architecture is a basic architecture for 5G-based metaverse applications. Despite the promise of latency and energy optimization, our proposed HMCC architecture needs to provide its adoption with application-specific requirements, such as data rates and real-time communication in terms of bandwidth limitation, coexistence with other cloud computing technologies, scalability, coverage, and security–for the future of IoT connectivity. We observe that machine learning based algorithms perform better than heuristic algorithms. When energy and latency constrained load balancing, DDPG-based algorithm gives better results. For networks with less edge servers but requiring lower latency, GRAPH algorithm proves useful.
Type of architecture
Proposed approaches
Limitations
Mobile application [
Mobile cloud computing [
Hybrid edge-cloud computing [
Our proposed hybrid mobile cloud computing (HMCC) architecture
Healthcare systems have certain constraints in terms of load computation, bandwidth, and security. Based on our literature survey, simulation of state-of-the-art algorithms for load balancing, observations, and our comparison results, we have noted some
In this paper, we proposed a hybrid mobile cloud computing (HMCC) architecture based on combined edge and cloud computing for healthcare applications. We designed it by keeping in mind concerns such as security, increasing traffic, latency, and energy consumption issues. Separation of public and private cloud ensures that vulnerable patient data is secured from the external public network. For managing network traffic for high throughput with minimal latency and energy consumption, we resorted to load balancing techniques. We compared static, as well as dynamic, load balancing algorithms with our architecture, and observed that dynamic algorithms provide better results. Dynamic algorithms, such as the graph coloring (GRAPH) algorithm, prove perfect when the number of edge servers in the network is less as compared to the workload, whereas the deep deterministic policy gradient (DDPG) algorithm is useful when both latency and energy conservation are of importance. We compare the results of these dynamic algorithms to the static algorithm, the energy efficient scheduling on federated edge cloud based on energy first (ESFEC-EF), and show better results are achieved using the dynamic approach. The edge cloud concept is a principal factor for the patients to gain access to the cloud from anywhere at any time. This architecture can form the basis for metaverse-based healthcare applications. Thus, our next goal is to provide a detailed implementation of HMCC architecture, and propose a dynamic load-balancing algorithm to support applications to function in the metaverse, and further adding blockchain technology to ensure secure transmission of encrypted patient data over the network [