Social big data management through collaborative mobile, regional, and cloud computing

Edge computing

Edge or fog computing is a recent world terminology aimed at extending cloud services to the network’s edge to reduce delay, response time, and costs. Edge computing (EC) is now massively used, especially in uncrewed vehicles and games (Ndikumana et al., 2021). Gu & Zhou (2019) proposed an algorithm, An Approximate Load Balancing Task Offloading Algorithm (ALBOA), to distribute vehicle workloads to different edge servers for minimum processing time. The experimental result shows that the algorithm achieves its objectives. Similarly, Chen et al. (2018) investigated resource-efficient edge issues for smart IoT applications. The proposed algorithm divides the smart application workload into similar nearby devices to minimize the delay. Their evaluation report stated that the proposed algorithm minimizes the delay of the service. Niu et al. (2019) explored the minimum service delay in Internet of Things (IoT) edge computing. Their algorithm divides the allocation into three different steps. First, the load balancing is carried out. Secondly, resources are allocated to the workload, and finally, converting the problem into an algorithm to solve it.

EC is useful in delay, response, and cost minimization. However, this domain needs to be expanded to MEC after the crowd of smart devices and technological shifts, making these devices practical for MEC.

Mobile computing

There is a crowd of smart devices around us. Fernando, Loke & Rahayu (2019) focused on this crowd and used nearby mobile devices to share the workload burden and not migrate it to edge or cloud computing. It saves energy and minimizes the cost and delay. Mobile devices use Bluetooth and WiFi to connect for data sharing. Fernando, Loke & Rahayu (2019) proposed a honeybee mobile computing structure using Bluetooth and WiFi for data sharing instead of cellular networks. The main challenge of the proposed structure lies in the limited battery power of Android devices. In the case of mobile computing, mainly when devices communicate directly, mobile collaboration is required, allowing mobile devices to respond to queries. The authors study the related concept in Mtibaa et al. (2013) to keep mobile energy consumption at a minimum, specifically in case the team is working in an emergency. Their proposed framework determines where to perform the task so that minimum energy can be consumed.

Direct sharing among mobile devices is an interesting idea; however, it has two main barriers. The first is the battery life, and the second is the users’ consent. Although the battery life is advancing, it still needs to be in a position to support continuous communication. Similarly, the standard operating procedures (SOPs) of direct communication need to be developed.

Hybrid environment

The collaborative structure of edge and cloud is usually used for delay, response, and cost minimization. Loghin, Ramapantulu & Teo (2019) explored the hybrid environment of cloud and edge computing and calculated the processing and transfer time on edge servers, cloud servers, and hybrid. Their comparison result shows that edge computing decreases the processing and transfer time.

Statistics show that about 46% of internet traffic is generated by mobile devices. To address these issues, Chen, He & Qiao (2019) proposed a mobile edge computing concept at the university level. The Universities’ infrastructure is used as servers, and the students’ mobiles use these resources.

The IoT devices produce big data. Large servers are required at the central processing unit to process the incoming data. Wan et al. (2019) used an uncrewed aerial vehicle-based station (UAV-BS), hovering at the sensors. The UAV-BS has sufficient capacity to process sensor data locally; however, when necessary, this data is transferred to the cloud server.

The most significant challenge in the cloud is communication delay. Despite efforts to minimize delay through job offloading, certain limitations persist. To tackle these challenges, the authors in Azure Speed Test (2021) suggested a collaboration between edge computing and cloud computing aimed at improving performance and lowering power consumption.

Likewise, in Ren et al. (2019), the authors suggested a collaboration between cloud and edge computing to tackle these challenges. They utilized the cloud for computing services and aimed at minimizing edge latency. Their findings indicate that the proposed approach effectively reduces latency in cloud communications. The evaluation results underscore that the collaboration of edge and cloud surpasses the individual use of either cloud or edge computing.

Mitigating the data overload from multimedia applications, a collaborative structure of mobile edge and cloud computing was proposed (Wu et al., 2017). To improve the user experience, they implemented strategies for device-to-device and machine-to-machine communication, effectively reducing the burden on base stations. Simulation results demonstrated a significant decrease in the workload on base stations.

A 30-frame video, having a resolution of 1024 × 1024, requires 125 TB of storage space when played 100 times per hour. Due to its high capacity requirements, Kioumourtzis et al. (2017) conducted research indicating that closed circuit television (CCTV) video streaming poses a risk of system-wide failure. In response, they introduced a hybrid cloud and edge computing concept, incorporating various video encoding techniques to takle this challenge. Experimental results demonstrated that the MPEG4 encoding scheme yielded a more reliable performance.

Operating in a hybrid environment poses several challenges. Firstly, specialized algorithms and technology are required to switch tasks between the edge and the cloud efficiently. Secondly, the necessity of transferring workloads to mobile computing has arisen with recent smart technology and the proliferation of smart devices. Therefore, this aspect requires further exploration.

Devices layer

This layer covers all devices connected to the communication system and the delay and cost caused by the data transfer from these devices to the base station, regional server, and cloud server.

Before explaining the working of the layer mentioned above, it is worth reporting here to explain the delays the communication faces (Bagubali et al., 2018). Four types of delay accumulate in communication. Transmission delay ( $D_{tran}$) is to transmit the data to the transmission media, propagation delay ( $D_{prop}$) is the data travel time on the transmission media from source to destination, queuing delay ( $D_q$) is the data waiting for the time in a queue to be processed by the system, and processing delay ( $D_{proc}$) when the system processes the data.

To compute the transmission delay ( $D_{tran}$),

(1) $$D_{\mathrm{t}\mathrm{r}\mathrm{a}\mathrm{n}}=\frac{W\times SNR\times ME}{B\times \mathrm{E}\mathrm{R}}$$where W is the workload (data), B is the channel bandwidth capacity, SNR is the signal-to-noise ratio (Zaman et al., 2023), modulation efficiency (ME) is the efficiency of the modulation scheme, and ER is the rate of transmission errors (Zaman et al., 2022).

The equation shows that the data transmission speed will also be high if the channel has a high bandwidth.

Similarly, to calculate the propagation delay ( $D_{prop}$),

(2) $$D_{prop}=\frac{Dis}{t{r}_s}$$where $Dis$ represents the distance, and $t{r}_s$ denotes the transmission speed.

To calculate the processing delay ( $D_{proc}$),

(3) $$D_{proc}=\frac{S}{P_r}$$where S is the size of the data and $P_{pr}$ is the processing rate of the particular machine.

Likewise, when the arrival of data increases the transmission or processing capacity, the data has to wait in the queue. The queuing delay ( $D_q$) depends on congestion.

The queuing delay is calculated as

(4) $$D_q=\frac{L\times a}{R}$$where L is the packet length, $a$ is the arrival rate of packets, and R is the packet processing rate.

The delay in mobile communication also depends on the devices connected to the base station; therefore,

(5) $$D_{MCI}=\frac{BSC}{NMC}$$

$D_{MC}$ is the delay caused by mobile connections impact, BSC is a base station capacity and NMC is the number of connected devices.

Therefore, the total delay faced by the data is calculated as;

(6) $$D_t=D_{tran}+D_{prop}+D_{proc}+D_{que}+D_{MCI}$$where $D_t$ shows the total delay data faces to reach the destination. $D_{tran}$ is the transmission delay, referring to the time from the transfer of the first bit to the last bit to the transmission medium. $D_{prop}$ shows the propagation delay; as the name refers, it shows the delay faced when the data propagates through the network. Similarly, $D_{proc}$ shows the data processing delay, and $D_{que}$ shows the queuing delay. For this study, the transmission delay ( $D_{tran}$) and propagation delay ( $D_{prop}$) are the key factors affecting the overall delay.

The layers discussed in this study use different channels and have different distances between users and servers. So, the delay will change according to the scenario.

Cellular communication layer

This article introduces this layer to share large files (i.e., images and videos) through BS station instead of streaming data from cloud servers and regional servers. It not only minimizes the delay and response time, but it also minimizes the operational cost.

The mobile devices delay is computed as follows:

(7) $$D{M}_t=2\times D_{tran}+2\times D_{prop}+D_q+D_{proc}$$

$D{M}_t$ shows the total delay of interaction with the base station, while $D_{tran}$ represents the transmission delay for transmitting the data. Similarly, $D_{prop}$ is the propagation delay between the mobile and base station, $D_q$ is the queuing delay, and $D_{Proc}$ is the processing delay at the base station. Propagation delay is minimum at MEC however, we have included it due to the wireless delay. In the case of mobile communication with BS, the queuing delay ( $D_q$) needs to be addressed because there is no masses connection with BS.

Algorithm 1 receives a content request from the users, and the optimum content location is suggested to stream the data to the target device with minimum time and cost. The base station keeps a list of the smart devices connected to it. The content is searched on these devices’ shared content. If the search is successful, the data is streamed to the target device; in the event of failure, the request is forwarded to the regional server.

The experimental cellular communication delay for ping is 12 ms. According to the research, this must stay the same by 20 ms.

The smart device layer encompasses all devices connected to the social network, including computers, laptops, smartphones, tablets, AI devices, and isolated devices. The communication layer consists of mobile networks that connect smart devices to edge computing. The anticipated 5G and 6G technologies provide the capability to transfer data instantly and without delay between devices and edge servers.

Regional layer

The Regional Computing server serves as the backbone of the proposed structure. It employs algorithms and AI technology to determine which content to store on the Edge Server or transfer to the cloud. The content is filtered using Algorithm 1. Content deemed unnecessary in another region is neither transferred to the cloud nor processed and stored locally. However, this data is transferred to the cloud servers during off-peak hours.

The delay faced by the regional server is calculated as follows:

(8) $$D{R}_t=D_{me}+D_{er}$$

$D{R}_t$ is the total delay faced while interacting with regional servers, and $D_{me}$ is the delay between mobile device and edge. Similarly, $D_{er}$ is the delay between edge and regional servers.

Similarly, the complexity of the regional servers depends on the total capacity of the regional server and the total workload on it.

(9) $$\mathrm{R}\mathrm{C}=\frac{CL}{MCC}$$

RC is the regional computing complexity, CL is the computational load, and MCC is the maximum computational capacity.

Algorithm 1 receives a content request from the users, and the optimum content location is suggested to stream the data to the target device with minimum time and cost. This algorithm receives the search request on the failed search at the base station. The content is searched on the regional servers. Upon a successful search, the data is streamed to the target devices, and in case of failure, the request is forwarded to the cloud servers.

Cloud layer

Cloud computing is the umbrella layer in this system, responsible for storing all data. However, the majority of computing services are distributed across the regional servers to minimize the data forwarded to the cloud for processing. The cloud servers are tasked with storing all users’ data, which is uploaded during off-peak hours.

The delay faced by the cloud is calculated as follows:

(10) $$D{C}_t=D_{me}+D_{er}+D_{rc}$$where $D{C}_t$ is the total delay faced while interacting with cloud servers, $D_{me}$ is the delay between mobile device and edge, $D_{er}$ is the delay between edge and regional servers and similarly the $D_{rc}$ is the delay between regional servers and cloud servers.

The major problem with the cloud is that traffic around the world connects to it, which increases the queuing delay ( $D_q$); similarly, due to the overload, it also increases the processing delay ( $D_{proc}$). Furthermore, due to the high delay, the transmitter devices consider it a packet loss and transmit another, creating an avalanche effect. Processing and storing the data nearby is necessary instead of transmitting it to distanced, congested servers to overcome the avalanche effect. In cellular communication, a delay up to 50 ms is considered good quality; however, once the delay increases from 200 ms, it has terrible effects on communication (Kumar, Manjunath & Kuri, 2008).

Algorithm 1 receives a content request from the users, and the optimum content location is suggested to stream the data to the target device with minimum time and cost. This algorithm receives the search request on the failure of a search at regional servers. The content is searched on the cloud servers. If the search is successful, the data is streamed to the target devices; in the case of failure, the request is forwarded to the regional servers and the base station.

Energy calculation

Likewise, to delay, energy consumption is also directly proportional to distance. More energy means more operational costs for data transfer. The energy consumption in the cloud environment is calculated as follows:

(11) $$E_{\mathrm{t}\mathrm{r}\mathrm{a}\mathrm{n}}=\sum _{j=1}^n{E}_{\mathrm{t}\mathrm{r}\mathrm{a}\mathrm{n}}(j,j+1)$$

$E_{tran}(j,j+1)$ represents the power consumption between consecutive stages. The power consumption increases as the stages increase.

where

(12) $$E_{tran}(j,j+1)=\frac{D_{j,j+1}\cdot P_j}{T_{j,j+1}}$$where $D_{j,j+1}$ is the distance, $P_j$ is the power and $T_{j,j+1}$ is the time.

(13) $$E_{other}=E_{pro}+E_{stor}+E_{col}+k.$$

Similarly, $E_{other}$ shows energy consumption in other activities such as energy consumption in processing ( $E_{pro}$), storing ( $E_{stor}$), and cooling the data centers ( $E_{col}$) (uz Zaman et al., 2019).

Therefore, the total energy utilization of communication is calculated as follows:

(14) $$E_{total}=E_{tran}+E_{other}.$$

Here, $E_{total}$ represents the total power usage, while $E_{tran}$ denotes the total energy usage during data transfer, encompassing wires, switches, routers, and other devices.

It is also known that,

(15) $$Cos{t}_{oper}\propto E.$$

The power consumption (E) is directly proportional to the operational cost ( $Cos{t}_{oper}$); consequently, higher power consumption leads to an increase in operational costs.

Experimental setup

The investigation is divided into three scenarios in the experimental setup, as shown in Fig. 3.

• The first scenario covers cloud servers and social media users worldwide.

• The second scenario covers the regional server, where the servers are established in a particular region, and the social media users directly access this server.

• The third scenario covers mobile edge computing, where mobile devices directly entertain the requests through the base stations.

The Cloud Analytic simulator do not support mobile-level communication. Therefore, the CISCO packet tracer is used with the Cloud Analytic to compute mobile device delay and response time.

Results analysis

The delay across the three computing paradigms shows a notable trend of increasing latency from MEC to RC and CC, as shown in Fig. 4. MEC consistently maintains the lowest delay at 12 ms, highlighting its efficient edge processing capabilities. In contrast, RC maintains a slightly higher but relatively low delay, ranging between 49 and 50 ms. However, CC exhibits the highest delay, with values fluctuating between 50 and 499 ms, signifying more variable and potentially slower processing times in the cloud. This observed progression underscores the trade-off between proximity and processing power. MEC offers the lowest latency due to its proximity to end devices, while CC, situated farther away, incurs the highest delay.

The cost analysis reveals distinct patterns among the three computing paradigms, as illustrated in Fig. 5. CC experiences significant variability in transfer costs, ranging from 0.065 to 0.228, with occasional spikes, indicating potential fluctuations and unpredictability in costs. In contrast, RC and MEC maintain consistent and low transfer costs, with RC at 0.065 and MEC at 0.02. This comparison underscores that, while CC may entail higher variability and costs, RC and MEC provide more stable and cost-efficient data transfer solutions. This emphasizes the advantages of leveraging regional or edge computing, particularly for applications with stringent cost considerations.

Figure 6 depicts the service level agreement (SLA) violation in milliseconds (ms) for different computing paradigms, which is calculated based on delay (Badshah et al., 2020b).

MEC outperforms CC and RC regarding SLA violations, providing consistently low delays and effectively meeting SLA. CC exhibits higher delay and, consequently, more SLA violations, while regional computing (RC) falls near MEC but still shows a higher delay than MEC. These results underscore the potential advantages of leveraging MEC for obtaining social big data with low latency.

The results presented in Fig. 7 indicate that cloud computing experienced penalties attributable to fluctuations in SLA violations, whereas EC consistently maintained a more stable SLA performance. MEC displayed outstanding SLA compliance, with no recorded penalties for SLA violations across all trials. These findings emphasize the considerable potential of MEC as an exceptionally efficient method for adhering to SLA without incurring extra penalties, thus reinforcing its viability for acquiring social big data.

Table 5 shows a comprehensive overview of the performance metrics for different computing paradigms. On average, cloud computing exhibits a delay of 275 ms, the highest among the paradigms, accompanied by a cost of $0.178, a 3.67% SLA violation rate, and penalties amounting to 1.83. In contrast, RC showcases a significantly lower average delay of 50 ms, a lower cost of $0.065, a minimal SLA violation rate of 0.83%, and lower penalties totaling 0.42. MEC outperforms both with an average delay of 12 ms, minimal cost ($0.02), no SLA violations, and zero penalties, highlighting its efficiency and reliability.

Comparative analysis

Numerous research endeavors have delved into the potential of edge computing to mitigate latency across diverse contexts. The article by Kuang et al. (2021) introduces an algorithm that harmonizes cooperative computation offloading, also resource allocation within the domain of mobile edge computing with the objective of latency reduction. Undoubtedly, edge computing brings advantages in terms of diminished delay and cost-effectiveness. However, it is crucial to underscore that a fundamental challenge lies in scalability, as edge computing inherently grapples with limitations.

Similarly, Zarandi & Tabassum (2021) presents a strategy to diminish latency within a multi-cell mobile edge computing network. This strategy involves optimizing offloading decisions and resource allocation. Nonetheless, this article contends with the formidable issue of scalability. Furthermore, in both Li, Deng & Deng (2021) and Zeng et al. (2020), authors employ comparable methodologies that leverage the synergy of edge computing and mobile computing to address the reduction of communication latency and transfer costs. However, the central concern prevails—the scalability dilemmas affiliated with edge and mobile computing persist as noteworthy challenges deserving further in-depth examination.

The scenario changes when considering social big data. Here, data requirements exhibit variation. In certain instances, data necessitates processing solely at the base station level, while in others, it requires elevated processing at the cloud level. Moreover, the priority of data migration assumes significance. Scenarios demand instantaneous data transfer to the cloud, while in other cases, there is flexibility to delay data migration until off-peak hours. Notably, the proposed collaboration encompassing mobile, regional, and cloud computing addresses issues that have yet to receive adequate attention within the existing literature.