Optimization of Location Allocation of Web Services Using a Modified Non-dominated Sorting Genetic Algorithm
Summary (3 min read)
1 Introduction
- Web Services are considered as self-contained, self-describing, modular applications that can be published, located, and invoked across the Web [25].
- With the ever increasing number of functional similar web services being available on the Internet, the web service providers (WSPs) are trying to improve the quality of service (QoS) to become competitive in the market.
- This algorithm solves the problem with one objective and one constraint.
- Among numerous EMO algorithms, Non-dominated sorting GA (NSGA-II) [4], Strength Pareto Evolutionary Algorithm 2 (SPEA-2) [5] have become standard approaches.
- – To model the web service location-allocation problem so that it can be tackled by NSGA-II, also known as The main objectives are.
2 Background
- In a GA, a population of strings (called chromosomes or the genotype of the genome), which are encoded as candidate solutions (called individuals, creatures, or phenotypes) to an optimization problem, evolves towards better solutions.
- Since a chromosome from the population represents a solution, when the algorithm starts, the whole population moves like one group towards an optimal area so the GA searches from a population of solutions rather than a single solution.
- Integer scalarization technique [3] is used to solve multi-objective problems with GA.
- When used for problems with only two objectives, NSGA-II performs relatively well in both convergence and computing speed.
- Parents are selected from the population by using tournament selection based on the rank and the crowding distance.
3.1 Problem Description and Assumptions
- Web service location-allocation problem is to determine reasonable locations for web services so that the deployment cost of WSP can be minimized while service performance can be optimized.
- In addition to deciding locations to of the services, information about network latency between user centers and candidate locations are needed.
- Therefore, the average network latency for a period of time should be representative.
- The details of these input data and modeling are introduced in Section 3.2.
- The authors made an assumption that WSPs periodicaly change the web service deployment since the user centers are changing over time.
3.2 Model Formulation
- For service location-allocation problem, the authors need information of service usage, network latency, and service deployment cost to decide service location-allocation so that the overall network latency can be minimized with minimal deployment cost and within constraints.
- The authors will use the following matrices.
- The network latency between user center i2 with candidate location j1 is 5.776s.
- These data could be collected by monitoring network latencies [31] [32].
- If there is more than one location, then the smallest latency is selected.
4.1 Chromosome Representation and Constraints
- The constraint setting is based on service providers’ needs.
- One can set multiply constraints to the problem to narrow the potential searching space.
- In their case, the authors set two basic constraints.
- The first constraint service number constraint requires that each service is deployed in at least one location.
- ∑ x∈S axj ≥ 1 (2) The second constraint, is cost constraint, which sets up the upper boundary of the total cost.
4.2 Genetic Operators
- The original NSGA-II uses a simulated binary crossover (SBX) [2] and polynomial mutation [24] to cope with continuous problems.
- The crossover point is created randomly within the length of the chromosome.
- In order to accomplish these two objectives, the authors design two fitness functions to evaluate how good each chromosome meets the objectives.
- The authors assume the latency is symmetrical between user center and candidate location.
- CostF itness′ = CostF itness− Costmin Costmax − Costmin (6) LatencyF itness′ = LatencyF itness− Latencymin Latencymax − Latencymin (7).
4.3 NSGA-II based algorithm for service location-allocation
- Comparing with the original NSGA-II their proposed algorithm has two new features.the authors.
- Firstly, in order to avoid repeatly evaluating the fitness of chromosomes, after the first generation is initialized, it storing the Pareto front in the memory.
- The reason for setting a memory pool is that, after a number of iterations, the population start converging.
- Secondly, it uses general mutation and crossover operation instead of polynomial mutation and simulated binary crossover.
- Therefore, repair operators are needed to try to maintain feasible solutions.
5 GA for Web Service Location Allocation
- In order to show the advantages of their multi-objective NSGA-II based approach, the authors extend the single-objective GA based approach in [13] to consider two objectives.
- The authors employ integer scalarization technique [9] to transform the multi-objective problem into a single objective problem.
- In their approach, the authors define the weight equals 0.5 since they consider both objectives equally important.
- Crossover and mutation operators are same as defined in Section 4.2.
- Note that CostFitness and LatencyFitness are calculated using Formula 6 and 7 in section 4.2.
6 Experiment Evaluation
- To evaluate the effectiveness and efficiency of their proposed NSGA-II based approach to service location-allocation.
- The authors compare their approach with the GA-based single objective approach in Section 5 using an existing dataset, WS-DREAM [31] [32], which is a historical dataset on QoS of Web services from different locations.
- In each dataset, algorithms are run under four different levels of cost constraints: Sufficient condition (indicating 100% the expected total cost), good condition (70%), pool condition (40%) and minimum budget condition (0%).
- In addition to the comparison between NSGA-II based algorithm and GA based algorithm, the authors conducted full experiments on NSGA-II with an initialisation and GA with an initialisation.
- The authors initialized the population with 49 random chromosomes and a chromosome which leading to minimum cost.
6.1 Effectiveness comparison
- The authors conducted experiments on NSGA-II, GA, NSGA-II with initialisation and GA with initialisation respectively.
- The authors goal is to minimize both cost and latency.
- The authors also notice that even though the population size is small as 50, including a chromosome of optimal cost can help to narrow down searching space and to converge to optimal solution faster.
- The results from initialized algorithms are similar with uninitialized algorithms, therefore the authors only present the uninitialized results.
7 Conclusion
- The authors proposed a NSGA-II based approach to web service location-allocation problem.
- The authors approach consider two objectives, minimizing cost and minimizing network latency at the same time.
- The authors have conducted a full experimental evaluation using the public WS-DREAM dataset to compare their approach to single-objective GA-based approach.
- The experimental results shows the NSGA-II based approach is effective to produce a set near-optima solutions for the web service location-allocation problem.
Did you find this useful? Give us your feedback
Citations
33 citations
Cites methods or result from "Optimization of Location Allocation..."
...Lastly, we conduct an experiment considering the overall performance of a BMOPSOCDwith a dynamic rounding function in comparison with three other algorithms: PSO, BNSPSO andNSGA-II (see [12])....
[...]
...compare its performance with BNSPSO, BPSO from our previous research [11], and NSGA-II from [12]....
[...]
...linear normalization method described in [12]....
[...]
...Therefore, we further investigated a Pareto front approach [12] with Sorting Genetic Algorithm-II (NSGA-II)....
[...]
...In contrast, in our previous studies [11], [12], we proposed a multi-objective algorithm with linear aggregation...
[...]
18 citations
9 citations
Cites background or methods from "Optimization of Location Allocation..."
...The network latency matrix [7] L = [luv] is used to model the round trip latency between each user group location and the location of a candidate VM, where luv is a float that indicates the number of seconds of the network latency....
[...]
...An NSGAII based approach is proposed in [7]....
[...]
...[7], [5] treat the problem of application deployment as location allocation problem with an aggregated objective that minimises the total cost and latency....
[...]
...Previous works have employed these techniques in the context of cloud deployment, however they either fail to consider multiple clouds [6] or neglect the use of local search to further optimise solutions [5], [7]....
[...]
7 citations
Cites methods from "Optimization of Location Allocation..."
...Due to its powerfulness of finding wide-spread solutions and implementation simplicity [13], NSGA-II has been successfully applied in many real-world multi-objective combinatorial problems such as web service allocation [14], service composition [15], [16] and resource allocation in clouds [17], [18]....
[...]
6 citations
References
305 citations
264 citations
"Optimization of Location Allocation..." refers methods in this paper
...The original NSGA-II uses a simulated binary crossover (SBX) [2] and polynomial mutation [24] to cope with continuous problems....
[...]
186 citations
184 citations
"Optimization of Location Allocation..." refers methods in this paper
...We employ integer scalarization technique [9] to transform the multi-objective problem into a single objective problem....
[...]
Related Papers (5)
Frequently Asked Questions (2)
Q2. What are the future works mentioned in the paper "Optimization of location allocation of web services using a modified non-dominated sorting genetic algorithm" ?
Future work will investigate the scalability of their proposed approaches for big datasets.