Performance evaluation of elephant flow predictors in data center networking

Bezerra, José Jailson Lima; Pinheiro, Antonio J.; Souza, Críston Pereira de; Campelo, Divanilson R.

doi:10.1016/j.future.2019.09.031

Cited by 12 publications

(10 citation statements)

References 27 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Owing to the ever-increasing number of cloud service users, DCNs have become a fundamental infrastructure for hosting different applications, such as web services, big data analytics, and cloud applications [8]- [10]. Moreover, current DCNs (e.g., Google cloud and Amazon) should meet the Quality of Service (QoS) requirements.…”

Section: Introductionmentioning

confidence: 99%

“…Much research [2], [10], [13], [16]- [21] has investigated DCN traffic and related flow management challenges either in SDN or other networking paradigms. Most of the proposed solutions in the DCN context take advantage of leveraging the SDN centralized control and the path diversity typically available in such networks.…”

Section: Introductionmentioning

confidence: 99%

“…Taking typical DCN flow types into consideration can reduce the shortcomings of current traffic management methods. Statistical studies conducted in DCNs [10], [22], [23] specify two typical flow types, namely Mice Flows (MFs) and Elephant Flows (EFs). MFs comprise 80% of the total flows in DCNs (e.g., emails and web pages) and are less than 100 KB in size.…”

Section: Introductionmentioning

confidence: 99%

“…On the other hand, EFs make up 10% of the total flows in DCNs (e.g., VM migrations and MapReduce) and are more than 100 KB in size. EFs are also long-lived (lasting minutes or more) and throughput-sensitive [10], [23], [24].…”

Section: Introductionmentioning

confidence: 99%

“…A concentration of EFs in DCNs can lead to a load imbalance because the Equal-Cost Multi-Path (ECMP) routing maps them to the same path. As a result, the load imbalance affects other flows attending for the utilization of the shared link [10]. To propose forwarding strategies for different flow types, some researchers concentrate on flow detection, particularly for EFs in an SDN environment [22].…”

Section: Introductionmentioning

confidence: 99%

See 4 more Smart Citations

Flow-Aware Forwarding in SDN Datacenters Using a Knapsack-PSO-Based Solution

Abdollahi

Deldari

Asadi

et al. 2021

IEEE Trans. Netw. Serv. Manage.

View full text Add to dashboard Cite

With the rapid growth of different massive applications and parallel flow requests in Data Center Networks (DCNs), today's providers are confronting challenges in flow forwarding decisions. Since Software Defined Networking (SDN) provides fine granular control, it can be intelligently programmed to distinguish between flow requirements. The present article proposes a knapsack model in which the link bandwidth and incoming flows are modeled as a knapsack capacity and items, respectively. Furthermore, each flow consists of two size and value aspects, acquired through flow size extraction and the type of service value assigned by the SDN controller decision. Indeed, the current work splits the incoming flow size range into Type of Service (ToS) decimal value numbers. The lower the flow size category, the higher the value dedicated to the flow. Particle Swarm Optimization (PSO) optimizes the knapsack problem and first forwards the selected-flows by KP-PSO, and the non-selectedflows second. To address the shortcomings of these methods in the event of dense parallel flow detection, the present study puts the link under the threshold of a 70% load by simultaneous requests. Experimental results indicate that the proposed method outperforms Sonum, Hedera, and ECMP in terms of flow completion time, packet loss rate, and goodput regarding flow size requirements.

show abstract

Section: Introductionmentioning

confidence: 99%