Abstract:Explicit Congestion Notification (ECN) is a deployment to IP and TCP also playing a crucial role in the congestion control of the Data Center Networks (DCNs). Most DCNs use a single queue scenario in each switch port. However, in the production of DCNs, the industry trend is moving towards one farther queue per port. Therefore, Multi-Service Multi-Queue Data Centers (MQ-ECN) have been proposed to ignite this service; afterward, DemePro improved the MQ-ECN. Yet, the fact that overhead and imprecise measurement are non-negligible for both models should be borne in mind. Also, MQ-ECN works for a round base scheduler while the overflow problem could be contributed by DemePro. Moreover, ECN was designed for single queue scenarios and having MQ-ECN is harmful at least for scheduling of flows. To solve the problem, we could take advantage of the lack of MQ-ECN and propose a machine-learning based dynamic threshold control scheme for ECN marking in DCN, which we named it DC-ECN (Data Center-Explicit Congestion Notification)–a first systematic solution to the problems which have already been mentioned earlier. The main point of DC-ECN is a separation of the mice and elephant flows in dual couple queues using machine learning. Then, to locate them into the requested queue with demand ECN marking threshold independently to achieve low latency and high throughput. Also, by dynamically increase and decrease the ECN marking threshold in elephant buffer, DC-ECN will never mark mice flows and succeed to absorb micro-burst mice traffic to have lowest latency without having sacrifice the throughput. Our mathematical analysis and simulation demonstrate that a steady state behavior of DC-ECN achieves 21.8% and 16.5% less flow completion time compared with MQ-ECN and DemoPro, respectively.

Priority Policy in Multi-Queue Data Center Networks Via Per-Port ECN Marking

Support ECN in Multi-Queue Datacenter Networks via Per-Port Marking with Selective Blindness

Improving ECN Marking Scheme with Micro-burst Traffic in Data Center Networks

Research on some active queue management algorithms

Early Marking for Controllable Maximum Queue Length in Data Center Networks.

DC-ECN: A machine-learning based dynamic threshold control scheme for ECN marking in DCN

One More Queue Is Enough: Minimizing Flow Completion Time With Explicit Priority Notification

ECN Marking with Micro-Burst Traffic: Problem, Analysis, and Improvement.

Centralized Congestion Control and Scheduling in a Datacenter

DiffECN: Differential ECN Marking for Datacenter Networks

Clean: Minimize Switch Queue Length Via Transparent ECN-proxy in Campus Networks

Alleviating Congestion via Switch Design for Fair Buffer Allocation in Datacenters

Re-architecting Congestion Management in Lossless Ethernet.

Poche: A Priority-Based Flow-Aware In-Network Caching Scheme in Data Center Networks

A New Rate-Based Active Queue Management Scheme over Wired-cum-Wireless Networks

Optimizing Flow Completion Time Via Adaptive Buffer Management in Data Center Networks

Design and Implementation of Fair Congestion Control for Data Centers Networks

Explicit Multipath Congestion Control for Data Center Networks

Congestion Control in Converged Ethernet with Heterogeneous and Time-Varying Delays.

DCCS: A dual congestion control signals based TCP for datacenter networks

A New Queue Discipline for Reducing Bufferbloat Effects in HetNet Concurrent Multipath Transfer