Application of a Long Short Term Memory neural predictor with asymmetric loss function for the resource allocation in NFV network architectures

Eramo, V.; Lavacca, Francesco G.; Catena, Tiziana; Salazar, Paul Jaime Perez

doi:10.1016/j.comnet.2021.108104

Cited by 28 publications

(25 citation statements)

References 37 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The performance of the APRPA algorithm is compared to the ones of two benchmark algorithms: (i) SPRPA in which the allocation is performed by applying a traditional approach with the objective to exactly predict the needed processing capacity and based on the minimization of the Root Mean Squared Error; (ii) ATPA [5,6] in which the resources are allocated on the basis of a traffic prediction based on both resource allocation and QoS degradation costs.…”

Section: Numerical Resultsmentioning

confidence: 99%

“…We have proposed the ATPA algorithm based on Seasonal Auto-Regressive Integrated Moving Average (SARIMA) [5,6] prediction with asymmetric loss function; because estimation errors are inevitable and because the error sign can have an impact on the network cost depending on the resource allocation and QoS degradation costs, we have proposed a solution in which the defined loss function gives higher (lower) weight to errors that result in a higher (lower) network cost. We have shown how the proposed solution allows for 30% network cost reduction with respect to the case in which a symmetric cost function as RMSE is chosen.…”

Section: Related Work and Research Contributionmentioning

confidence: 99%

“…Finally the APRPA algorithm is compared with our previous ATPA solution [5,6] in Figure 9 in which traffic prediction is only performed. We can observe the better performance of APRPA with 10% gain.…”

Section: Number Of Allocatedmentioning

confidence: 99%

“…We have proposed the Asymmetric Traffic Prediction-based Allocation (ATPA) algorithm [5][6][7] in which the resources are allocated on the basis of a traffic prediction based on both resource allocation and QoS degradation costs. The allocation cost is the rent cost of the processing cores.…”

Section: Introductionmentioning

confidence: 99%

“…The only minimization of the RMSE may lead to worse performance because of the error sign that differently impacts on the total network cost when the allocation and QoS degradation costs are different.In particular we can notice from Figure6how the increase in C QoS leads to a rapid increase in the total network cost in the SPRPA solution because it is not able to limit the total QoS degradation cost as highlighted in Figure8. Conversely the APRPA solution is able to apply the appropriate countermeasures, to over-allocate the processing resource and to reduce the total QoS degradation cost as highlighted in Figures7 and 8where we notice an increase in total resource allocation cost (due to the resource over-allocation) and a decrease in total QoS degradation cost.Finally the APRPA algorithm is compared with our previous ATPA solution[5,6] in Figure9in which traffic prediction is only performed. We can observe the better performance of APRPA with 10% gain.…”

mentioning

confidence: 99%

See 4 more Smart Citations

Proposal and Investigation of a Convolutional and LSTM Neural Network for the Cost-Aware Resource Prediction in Softwarized Networks

Eramo¹,

Valente²,

Catena³

et al. 2021

Future Internet

Self Cite

View full text Add to dashboard Cite

Resource prediction algorithms have been recently proposed in Network Function Virtualization architectures. A prediction-based resource allocation is characterized by higher operation costs due to: (i) Resource underestimate that leads to quality of service degradation; (ii) used cloud resource over allocation when a resource overestimate occurs. To reduce such a cost, we propose a cost-aware prediction algorithm able to minimize the sum of the two cost components. The proposed prediction solution is based on a convolutional and Long Short Term Memory neural network to handle the spatial and temporal correlations of the need processing capacities. We compare in a real network and traffic scenario the proposed technique to a traditional one in which the aim is to exactly predict the needed processing capacity. We show how the proposed solution allows for cost advantages in the order of 20%.

show abstract

Section: Numerical Resultsmentioning

confidence: 99%

Section: Related Work and Research Contributionmentioning

confidence: 99%

Section: Number Of Allocatedmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

mentioning

confidence: 99%

See 3 more Smart Citations

Proposal and Investigation of a Convolutional and LSTM Neural Network for the Cost-Aware Resource Prediction in Softwarized Networks

Eramo¹,

Valente²,

Catena³

et al. 2021

Future Internet

Self Cite

View full text Add to dashboard Cite

show abstract

Dynamic service function chains placement based on parallelized requests in edge computing environment

Guo,

Rezaeipanah

2023

Trans Emerging Tel Tech

View full text Add to dashboard Cite

The advent of Network Function Virtualization (NFV) technology brings flexible traffic engineering to edge computing environments. Online services in NFV are chained as Service Function Chains (SFCs), which consist of ordered sequences of Virtual Network Functions (VNFs). The SFC Placement (SFCP) problem is solved under Quality of Service (QoS) requirements and limited resource availability by directing traffic to the required VNFs. However, SFC assembly leads to high latency and network congestion by increasing the count of VNFs, which parallelized SFC can overcome this problem. With parallelizing an SFC request, independent VNFs are activated simultaneously and computational acceleration is realized by reducing the SFC length. Any pair of VNFs that do not conflict with traffic can be activated simultaneously. Most VNFs are deployed on distributed servers for load balancing, which makes SFC parallelization challenging. Meanwhile, the cost of merging/duplicating packets for parallelized SFCs between different servers is not negligible. Hence, in this article, Distributed Parallel Chaining (DPC) is proposed which is an algorithm based on Deep Reinforcement Learning (DRL) approaches. The DPC algorithm solves the SFCP problem to maximize the Long‐Term Expected Cumulative Reward (LTECR). DPC incorporates an Asynchronous Advantage Actor‐Critic (A3C) algorithm as a new approach based on DRL to increase the admission ability of future SFC requests by maximizing LTECR. The evaluation results show the effectiveness of the proposed algorithm from different aspects. Specifically, compared to the best existing approaches, DPC can reduce SFC latency by 8.7%.

show abstract

NFVLearn: A multi‐resource, long short‐term memory‐based virtual network function resource usage prediction architecture

2022

View full text Add to dashboard Cite

Virtual resource load prediction in network function virtualization (NFV) is the subject of intense research due to its crucial role in enabling proactive resource adaptation in dynamic NFV environments whose resource demand constantly changes. Several long short‐term memory (LSTM)‐based approaches have been proposed to forecast the resource load of multiple resource attributes of a virtual network function (VNF) in a service function chain (SFC). In this article, we present NFVLearn, a flexible multivariate, many‐to‐many LSTM‐based model which uses different types of resource load history (CPU, memory, I/O bandwidth) from various VNFs of an SFC to predict future loads of multiple resources of a VNF. We then compare four novel automated input selection frameworks for NFVLearn. Simulations on those frameworks based on graph neural networks, Pearson correlation coefficient, Spearman rank correlation coefficient, and Kendall rank correlation coefficient demonstrate that models using lesser, highly correlated input features retain high prediction root mean squared error accuracy and coefficients of determination scores by leveraging resource attribute inter‐dependencies from the SFC. Those results show that resource attribute interdependency‐based input feature selection frameworks can reduce overhead in the control plane while keeping high accuracy and high fidelity resource load prediction of multiple resource attributes.

show abstract

Application of a Long Short Term Memory neural predictor with asymmetric loss function for the resource allocation in NFV network architectures

Cited by 28 publications

References 37 publications

Proposal and Investigation of a Convolutional and LSTM Neural Network for the Cost-Aware Resource Prediction in Softwarized Networks

Proposal and Investigation of a Convolutional and LSTM Neural Network for the Cost-Aware Resource Prediction in Softwarized Networks

Dynamic service function chains placement based on parallelized requests in edge computing environment

NFVLearn: A multi‐resource, long short‐term memory‐based virtual network function resource usage prediction architecture

Contact Info

Product

Resources

About