Enhancing workflow efficiency with a modified Firefly Algorithm for hybrid cloud edge environments Ali Guerra | usagoldmines.com

On this part, we current an in depth evaluation of our experimental setup, efficiency metrics, obtained outcomes, and their complete evaluation. We conduct a radical evaluation of the outcomes, delving into the implications of our findings and uncovering developments and insights concerning the effectiveness and appropriateness of ModFOA for scientific workflow scheduling inside cloud computing environments.

Simulation setting and benchmark workflows

To guage the efficiency of our proposed algorithm, we carried out the ModFOA algorithm utilizing CloudSim, a widely known and extensively used simulation framework within the cloud computing analysis area31,33. CloudSim’s versatility and sturdy group assist rendered it a pretty and sensible alternative for our examine, providing the required flexibility to customise the simulator to our particular necessities. Notably, we integrated a Directed Acyclic Graph (DAG) into CloudSim to characterize advanced activity dependencies, making it well-suited for our investigation into multi-objective activity scheduling in cloud computing37. To evaluate the effectiveness of the ModFOA algorithm, we performed a comparative evaluation in opposition to a number of current activity scheduling algorithms, together with GA employed in38, ACO39, and Hybrid Genetic Algorithm (HGA) proposed in40.

To configure the experiment, we fastidiously outlined varied algorithm parameters similar to inhabitants dimension, iteration restrict, and the dynamic step dimension parameter (alpha) to control the trade-off between exploration and exploitation, as depicted in Desk 2. These parameters had been initialized in accordance with the actual calls for of the issue being investigated and insights derived from earlier experimentation.

The experimental setup commenced with a meticulous delineation of the issue, encompassing the traits of the scientific workflow, activity dependencies, useful resource constraints, and optimization goals. Algorithmic parameters had been subsequently fine-tuned to align with the issue specs, making certain that the ModFOA algorithm was suitably tailor-made for the duty at hand. The algorithm was then executed, capitalizing on its programming prowess to iteratively optimize the target operate in accordance with the predefined downside and parameter settings.

Publish-execution, efficiency analysis ensued to gauge varied metrics similar to convergence charge, resolution high quality, execution time, and useful resource utilization. MATLAB’s in-built visualization and evaluation instruments proved invaluable in deciphering the outcomes successfully. To bolster the reliability and robustness of our findings, a number of runs of the algorithm had been performed utilizing completely different random seeds or downside situations. Statistical evaluation methods had been subsequently employed to scrutinize the variability and consistency of the outcomes, thereby furnishing insights into the algorithm’s efficiency throughout various circumstances.

The experiments had been performed on a desktop pc outfitted with ample computational assets, together with an Intel Core i7 processor and 8GB RAM, to make sure environment friendly dealing with of simulations. This setup facilitated the execution and evaluation of experiments inside cheap timeframes, enabling a complete evaluation of the ModFOA algorithm’s efficacy in scientific workflow scheduling and optimization duties.

To make sure our simulation precisely displays communication and computation prices, we set up two key graphs: (GR1_{n occasions n}) and (GR2_{n occasions m}).

(GR1_{n occasions n}), often known as the communication value graph, simulates knowledge switch occasions between duties within the workflow. Each activity inside the Directed Acyclic Graph (DAG) symbolizing the workflow bears communication prices, randomly generated inside a span of 1 to 10 time models. These prices, allotted to edges inside the DAG, depict the interaction amongst duties.

(GR2_{n occasions m}), termed the computing value graph, characterizes activity execution occasions throughout completely different processors. Computing prices, reflecting processor capabilities and activity complexities, are chosen from a variety of 10 to 30 time models. These prices characterize various execution occasions influenced by components similar to useful resource capacities and activity intricacies. Derived from processing value values specified within the cloudlet info, computing prices regulate execution durations primarily based on the variety of accessible processors, which is about to five in our simulation. On this examine, the time period “cloudlet” refers to a smaller computational activity or unit of labor inside a bigger workflow that must be processed. Cloudlets are the person duties that make up an entire workflow, and they’re scheduled throughout accessible assets, similar to VMs in cloud or edge environments. The cloudlet’s attributes, similar to its computational necessities and knowledge dependencies, are essential for figuring out probably the most environment friendly method to allocate assets and reduce the general makespan of the workflow.

The chosen vary of 10 to 30 time models for computing prices caters to the execution of duties throughout 5 processors, making certain a sensible portrayal of assorted execution durations. This strategy enhances the credibility and accuracy of our simulation setting, essential for efficient scheduling choices in scientific workflow optimization duties.

Efficiency metrics

In evaluating the efficiency of scheduling algorithms for scientific workflows, a number of key metrics are generally used: makespan, flowtime, and price.

1.

Makespan: Makespan refers back to the whole time taken to execute all duties within the workflow from begin to end22,23. It stands as a pivotal metric for gauging the effectiveness of a scheduling algorithm in carrying out your entire workload. Mathematically, makespan ((C_{textual content {max}})) may be calculated because the completion time of the final activity within the workflow as per Eq. (14).

the place (C_i) represents the completion time of activity i.

2.

Flowtime: Flowtime denotes the period taken for a activity to finalize its execution from the moment it enters the system till its completion18. It supplies insights into the person activity execution occasions and may help establish bottlenecks within the workflow. The flowtime ((FT_i)) of activity i may be calculated as per Eq. (15).

the place (C_i) is the completion time of activity i and (r_i) is the discharge time of activity i.

3.

Value: Value is a metric that considers varied resource-related components, similar to vitality consumption or financial bills, related to executing the workflow24. The associated fee (Value) may be computed as a operate of makespan and different resource-related parameters. For instance, a easy value mannequin could also be outlined as per Eq. (16).

the place Useful resource Value represents the price incurred per unit time for using computational assets.

4.

Useful resource Utilization: It refers to how successfully a system makes use of all its assets collectively9. It affords a holistic perspective on the effectivity of useful resource utilization throughout your entire system. Combination useful resource utilization, denoted as RU, may be computed by contemplating the sum or common of particular person useful resource utilizations throughout all parts of the system. Let’s outline the person useful resource utilizations as follows:

One methodology to calculate mixture useful resource utilization is the common utilization, expressed as Eq. (17).

Right here, N represents the whole variety of parts within the system. By evaluating mixture useful resource utilization, we acquire insights into how effectively assets are utilized throughout the system as an entire. This perception is essential for optimizing useful resource allocation, figuring out bottlenecks, and enhancing general system efficiency.

5.

Power Consumption: It’s a vital efficiency metric in cloud computing, because it straight impacts operational prices and environmental sustainability27. It may be calculated because the sum of vitality consumed by all parts inside the cloud infrastructure, together with servers, networking gear, cooling methods, and storage units. The equation for whole vitality consumption ((EC)) in cloud computing may be expressed as Eq. (18).

The place: – (E_{textual content {servers}}) represents the vitality consumed by servers, together with each energetic and idle energy consumption. – (E_{textual content {networking}}) is the vitality consumed by networking gear, similar to switches and routers, throughout knowledge transmission and reception. – (E_{textual content {cooling}}) denotes the vitality expended by cooling methods, similar to air con models and followers, to keep up optimum working temperatures inside knowledge facilities. – (E_{textual content {storage}}) displays the vitality consumed by storage units, together with laborious disk drives (HDDs) and solid-state drives (SSDs), throughout knowledge learn and write operations. Environment friendly useful resource allocation, workload consolidation, and dynamic energy administration methods may help reduce whole vitality consumption in cloud environments, resulting in value financial savings and lowered carbon emissions.

These metrics present a complete evaluation of the efficiency of scheduling algorithms in scientific workflow optimization duties, contemplating each particular person activity execution traits and general workflow effectivity.

Comparative outcome and their evaluation

This part delves into assessing the efficiency of the proposed algorithm via the examination of assorted parameter combos. The analysis revolves round a health operate that encompasses three main goals: minimizing execution time (Makespan, (C_{textual content {max}})), optimizing movement time ((F)), and lowering value. Weights ((w_1), (w_2), and (w_3)) are allotted to every goal inside the health operate to suggest their relative significance. The exploration of those parameter combos unveils the trade-offs amongst these goals. Desk 3 presents variations in health weights for various schemes, every characterised by a mixture of weights assigned to a few efficiency metrics: Makespan, Flowtime, and Complete Value. Makespan refers back to the whole period required to finish all duties in a scheduling situation, whereas Flowtime represents the whole time spent by duties within the system from initiation to completion. Complete Value encompasses the general expenditure related to activity execution and useful resource utilization. Every scheme is denoted by a novel identifier, with corresponding weight values assigned to Makespan ((w_1)), Flowtime ((w_2)), and Complete Value ((w_3)). As an example, in Scheme 1, equal significance is given to Makespan and Flowtime, whereas Complete Value is prioritized with the next weight in Scheme 4. These variations in health weights allow the exploration of various trade-offs between Makespan, Flowtime, and Complete Value in activity scheduling and optimization eventualities.

To comprehensively consider the algorithm’s efficiency throughout varied useful resource configurations, we introduce a number of experimental setting in Desk 4. Desk 4 illustrates variations in useful resource configuration throughout completely different eventualities inside the cloud computing setting. Every configuration, labeled from Config 1 to Config 4, represents a definite mixture of the variety of cloudlets, hosts, and digital machines (VMs). Config 1 consists of a comparatively small-scale setup with 100 cloudlets distributed throughout 2 hosts, every internet hosting 5 VMs. In distinction, Config 4 represents a larger-scale deployment with 1000 cloudlets serviced by 20 hosts, every accommodating 150 VMs. Because the configurations progress from Config 1 to Config 4, there’s a noticeable enhance within the scale of the infrastructure, with Config 4 being probably the most resource-intensive setup. These variations in useful resource configurations enable for the examination of efficiency throughout completely different workload sizes and infrastructure capacities, offering insights into scalability and useful resource utilization effectivity inside cloud environments.

A sequence of experiments was carried out, exploring various configurations by assigning various weights to the ModFOA strategy and current algorithms. This comparative evaluation centered on makespan, useful resource utilization, and vitality consumption, as outlined in Part 6.2. The outcomes obtained from these experiments are elaborated upon and introduced within the subsequent subsections.

Makespan-based outcome evaluation

Determine 1 presents the Makespan outcomes obtained from varied configurations (as per Desk 4) of ModFOA utilized to completely different eventualities, doubtlessly representing distinct scientific workflows, by various health weights (as per Desk 3). The info highlights a variety of completion occasions, starting from 176,543 to 389,728, relying on the chosen configuration and situation. Whereas figuring out a single optimum ModFOA scheme primarily based solely on Makespan proves difficult, sure developments emerge. Schemes 1, 6, and 9 constantly exhibit greater completion occasions throughout configurations, suggesting potential limitations in minimizing workflow execution time. Conversely, Schemes 2, 7, and eight reveal promise with decrease Makespan values in a number of configurations.

Notably, inside a single scheme, Makespan reveals variability throughout configurations, indicating the affect of particular parameter settings on the algorithm’s scheduling effectivity. To additional enrich the evaluation, further particulars on every scheme’s configuration specifics (e.g., parameter values) and the variety of workflows per situation could be useful. Furthermore, incorporating knowledge on useful resource utilization and price metrics would supply a extra holistic analysis of the algorithm’s efficiency. In essence, this evaluation underscores the importance of judiciously choosing ModFOA configurations tailor-made to the distinctive attributes of scientific workflows and the specified efficiency goals, similar to minimizing completion time.

Determine 2 illustrates the makespan outcomes obtained from varied configurations (as outlined in Desk 4) of ModFOA and current approaches utilized to various eventualities, doubtlessly representing distinct scientific workflows, by adjusting health weights (as per Desk 3). Total, ModFOA demonstrates superior efficiency in comparison with ACO and GA throughout most configurations, suggesting its potential as a extra environment friendly alternative for these particular workflows. Nevertheless, notable variation in efficiency exists inside ModFOA, each between schemes and configurations, indicating the significance of optimizing parameter settings inside the chosen algorithm to additional improve efficiency. In distinction, ACO and GA constantly exhibit greater makespan values, indicating their doubtlessly lesser suitability for this particular set of workflows. PSO, nonetheless, exhibits aggressive efficiency, with makespan values corresponding to ModFOA and even attaining decrease completion occasions in sure configurations. Nonetheless, a single excessive makespan worth for PSO raises considerations concerning its consistency throughout completely different eventualities. To achieve deeper insights, understanding the precise configuration particulars for every scheme inside every algorithm could be useful. Moreover, info on the variety of workflows included in every scheme and different efficiency metrics similar to useful resource utilization or value would supply a extra complete evaluation.

The outcomes spotlight the effectiveness of ModFOA schemes in comparison with conventional algorithms (ACO, GA, PSO) throughout varied configurations. In Config 1, ModFOA schemes constantly outperform conventional strategies, with the best-performing ModFOA scheme, Scheme_7, attaining a worth of 12, which is a 56% enchancment over ACO (27) and a 40% enchancment over GA (20). This development continues in Config 2, the place ModFOA (Scheme_2) achieves a worth of 31, representing a 19% enchancment over ACO (38) and performing carefully with GA (35) and PSO (40).

In Config 3, ModFOA schemes present distinctive efficiency, with Scheme_10 attaining a worth of 47, marking a 7% enchancment over GA (49) and a 41% enchancment over PSO (80). Equally, in Config 4, ModFOA schemes like Scheme_10 and Scheme_6, each with values of 24, exhibit a 33% enchancment over PSO (36) and an 18% enchancment over GA (31). Total, the ModFOA schemes reveal constant and substantial enhancements throughout configurations, emphasizing their superiority in optimizing activity scheduling over conventional algorithms.

In conclusion, this evaluation underscores the importance of choosing the suitable algorithm and finely tuning its configuration for environment friendly scientific workflow scheduling. Whereas ModFOA reveals promise for these eventualities, additional investigation and optimization inside every algorithm are crucial to establish probably the most environment friendly configuration for a given workflow.

Useful resource usage-based outcome evaluation

Determine 3 presents the useful resource utilization percentages achieved by completely different configurations of the ModFOA algorithm throughout 4 experimental setups (Configs 1 to 4). The useful resource utilization metric represents the proportion of obtainable assets utilized through the execution of scientific workflows. Analyzing the outcomes reveals a number of developments. Firstly, there’s variation in useful resource utilization throughout completely different schemes of the ModFOA algorithm inside every configuration. This means that the precise parameter settings and optimization methods employed by every scheme affect useful resource utilization effectivity. Secondly, evaluating useful resource utilization throughout configurations, it’s evident that sure configurations result in greater useful resource utilization percentages. For instance, Configuration 3 constantly demonstrates greater useful resource utilization in comparison with different configurations throughout varied ModFOA schemes. This implies that the variety of cloudlets, hosts, and VMs allotted in Configuration 3 could also be higher suited to completely make the most of accessible assets. Moreover, some ModFOA schemes constantly exhibit decrease useful resource utilization percentages throughout all configurations. As an example, Scheme 2 constantly achieves decrease useful resource utilization in comparison with different schemes. This might point out that Scheme 2 employs optimization methods that end in extra environment friendly useful resource allocation and utilization. Total, the evaluation of useful resource utilization underscores the significance of configuring algorithm parameters and experimental setups appropriately to maximise useful resource utilization effectivity. Moreover, it highlights the potential for particular ModFOA schemes and configurations to attain extra optimum useful resource allocation, resulting in improved general efficiency in scientific workflow scheduling.

Determine 4 presents the useful resource utilization values for various configurations (Config 1, Config 2, Config 3, and Config 4) beneath varied schemes of ModFOA, ACO, GA, and PSO. These values characterize the proportion of useful resource utilization, indicating how effectively the algorithms allocate assets for executing scientific workflows. Analyzing the outcomes, we observe that ModFOA usually reveals greater useful resource utilization in comparison with ACO, GA, and PSO throughout most configurations and schemes. This implies that ModFOA effectively makes use of accessible assets for activity execution, main to higher general useful resource utilization. Inside ModFOA, there’s variability in useful resource utilization between completely different schemes and configurations. As an example, Scheme 7 constantly demonstrates greater useful resource utilization values in comparison with different schemes, indicating doubtlessly extra environment friendly useful resource allocation methods. Conversely, Scheme 2 and Scheme 4 exhibit decrease useful resource utilization values in some configurations, suggesting much less optimum useful resource allocation. In distinction, ACO and GA usually present decrease useful resource utilization values in comparison with ModFOA, indicating doubtlessly much less environment friendly useful resource allocation methods. PSO, nonetheless, demonstrates aggressive useful resource utilization values corresponding to ModFOA in some configurations, suggesting its effectiveness in useful resource allocation for sure eventualities.

The ModFOA schemes reveal substantial enhancements over conventional algorithms (ACO, GA, PSO) throughout the completely different configurations. In Config 1, ModFOA schemes present a big benefit, with Scheme_7 attaining a worth of 64, representing a formidable 88% enchancment over ACO (34) and a 156% enchancment over GA (25). In Config 2, ModFOA schemes preserve their superiority, with Scheme_3 reaching a worth of 92, indicating a 41% enchancment over ACO (65) and a 207% enchancment over GA (30). This development illustrates the effectiveness of ModFOA in outperforming conventional strategies throughout varied settings.

In Config 3, ModFOA schemes similar to Scheme_10 attain a worth of 98, showcasing a 2% enchancment over GA (92) and a 24% discount in comparison with PSO (130). In Config 4, ModFOA schemes once more excel, with Scheme_7 attaining a worth of 82, which represents a 49% enchancment over ACO (55) and a 32% enchancment over GA (62). These outcomes spotlight that ModFOA schemes constantly supply superior efficiency in comparison with ACO, GA, and PSO, demonstrating their effectiveness in optimizing activity scheduling with appreciable enhancements throughout completely different configurations.

Total, the evaluation of useful resource utilization highlights the significance of algorithm choice and scheme configuration in attaining environment friendly useful resource utilization for scientific workflow scheduling. Whereas ModFOA usually performs effectively in useful resource allocation, additional investigation into the precise schemes and configurations is important to establish probably the most environment friendly strategy for various eventualities.

Power consumption-based outcome evaluation

Determine 5 presents vitality consumption values for various schemes of ModFOA throughout various numbers of cloudlets. These values characterize the whole vitality consumed by the system through the execution of scientific workflows. Analyzing the outcomes, we observe that vitality consumption usually will increase with an rising variety of cloudlets for all schemes of ModFOA. That is anticipated as extra cloudlets require greater computational assets, resulting in elevated vitality consumption. Throughout completely different schemes, we discover variations in vitality consumption patterns. Some schemes, similar to Scheme 1 and Scheme 4, reveal comparatively constant vitality consumption values throughout completely different numbers of cloudlets, suggesting stability in vitality utilization no matter workload dimension. In distinction, schemes like Scheme 10 exhibit extra vital fluctuations in vitality consumption, indicating potential inefficiencies in useful resource allocation or activity scheduling methods. Moreover, there seems to be an general rising development in vitality consumption as we transfer from Scheme 1 to Scheme 10, indicating that sure scheme configurations might result in greater vitality consumption than others. This underscores the significance of fastidiously choosing and optimizing scheme parameters to attenuate vitality utilization whereas making certain environment friendly workflow execution. Total, the evaluation of vitality consumption highlights the necessity for optimizing scheme configurations and useful resource allocation methods to attenuate vitality utilization and enhance general system effectivity in scientific workflow scheduling. Additional investigation into the precise traits and efficiency implications of every scheme is important to establish probably the most energy-efficient strategy for various eventualities.

Determine 6 presents vitality consumption values for various schemes, together with ModFOA, ACO, GA, and PSO, throughout various numbers of cloudlets. These values characterize the whole vitality consumed by the system through the execution of scientific workflows. Analyzing the outcomes, we observe that vitality consumption varies considerably throughout completely different schemes and numbers of cloudlets. For ModFOA (Scheme 1), vitality consumption usually will increase because the variety of cloudlets will increase, which is anticipated as a result of greater computational load. Evaluating ModFOA (Scheme 1) with different algorithms, we discover that ACO tends to have greater vitality consumption values throughout most configurations. This implies that ACO might not be as environment friendly when it comes to vitality utilization in comparison with ModFOA. Equally, GA and PSO additionally exhibit comparatively excessive vitality consumption values, particularly because the variety of cloudlets will increase. This means that these algorithms might eat extra vitality to execute scientific workflows in comparison with ModFOA.

The efficiency of ModFOA throughout varied cloudlet configurations demonstrates notable enhancements over conventional algorithms (ACO, GA, PSO). In configurations with 100 cloudlets, ModFOA (Scheme_1) exhibits a big 70% enchancment over ACO (676) and a 44% enchancment over PSO (833). The efficiency benefit turns into much more pronounced with greater cloudlet counts, the place ModFOA achieves enhancements of 74% over GA (1118) and 116% over PSO (833) in 100 cloudlets, and these enhancements proceed to increase throughout subsequent configurations.

Because the variety of cloudlets will increase to 1000, ModFOA (Scheme_1) reaches a worth of 1800, showcasing a 16% enchancment over ACO (2143) and a 27% enchancment over PSO (2245). In comparison with GA, ModFOA delivers a 27% enchancment, reaching 1800 versus GA’s 2454. These enhancements spotlight ModFOA’s constant superiority in dealing with bigger cloudlet sizes, offering vital enhancements in efficiency metrics in comparison with ACO, GA, and PSO throughout all examined configurations.

The proposed strategy affords a number of key benefits over current algorithms. Firstly, it integrates dynamic step dimension adjustment, quasireflection-based studying, and opposition-based studying methods, which improve exploration and exploitation of the answer area, decreasing the chance of untimely convergence. Moreover, our algorithm successfully balances a number of goals similar to makespan, flowtime, and price, resulting in extra optimized useful resource utilization within the cloud-edge setting. These improvements end in improved efficiency, as demonstrated by the decrease vitality consumption and higher scheduling effectivity in our experimental outcomes.

On the whole, the evaluation highlights the significance of choosing an energy-efficient algorithm for scientific workflow scheduling. Though ModFOA (scheme 1) usually demonstrates reasonable vitality consumption, additional investigation is important to establish probably the most vitality environment friendly strategy for various eventualities. As well as, optimizing algorithm parameters and useful resource allocation methods might assist reduce vitality utilization and enhance general system effectivity.

Our proposed strategy has sure limitations. One limitation is that the effectiveness of the algorithm may be influenced by the number of parameters such because the step dimension and the randomization issue, which can require fine-tuning for various functions. Moreover, whereas our algorithm performs effectively in cloud-edge environments, its efficiency might range when utilized to extremely heterogeneous or large-scale distributed methods, and scalability could possibly be an space for future enchancment.