Research On Optimal Scheduling Strategy For High Performance Data Center Networks

With the continuous promotion of cloud computing technology,modern datacenters involve an ever-growing array of mixed workloads,each of which puts forward different service-level objective,including high throughput,low latency,etc.Currently,most datacenters introduce statistical multiplexing technology and oversubscription to the network design for lowering the total cost,which can easily lead to the occurrence of network congestion,high latency,low throughput and etc.Currently,most of datacenter network optimization solutions focus on solving problems at the protocol layer,but such solutions typically passively conduct countermeasures,which is inefficient.Other physical layer solutions are expensive which is hard-promoted,and some centralized arbitration solutions have also been widely criticized for their poor scalability.In view of the above problems,an optimal scheduling strategy for high performance data center networks based on time triggered coordination(Pro CAM)and an optimal scheduling strategy for high performance data center networks based on time/event triggering synthesis are proposed respectively in this paper.Pro CAM utilizes the predictability of throughput-intensive flows to construct an optimized low-congestion transmission model and finally achieves network congestion avoidance.Hybrid Fass classifies flows into Time-Triggered flows and Best-Effort flows,and constructs a globally minimized transmission delay model for Time-Triggered flows and implements a high-throughput transfer mechanism for Best-Effort flows.The main work and contributions of this paper are as follows:1)Based on the software defined network architecture,this paper designs working mechanisms of each module in Pro CAM architecture,constructs a lowcongestion transmission model with its fast solution algorithm,to prearrange optimal coordinate scheme(desynchronize the sending time of concurrent longflows)beforehand as a proactive manner,and proposes a dynamic bandwidth allocation strategy for predictable flows and normal flows.2)By discussing the packet transmission delay composition in the network,Hybrid Fass constructs a flow-level transmission delay optimization model,a task-level transmission delay optimization model and a system-level transmission delay optimization model for Time-Triggered flows.For other flows,Hybrid Fass proposes the Best-Effort transmission mechanism.3)Through the design of simulation scenarios and data,this paper verifies that Pro CAM can effectively avoid network congestion,Hybrid Fass can simultaneously achieve low latency transmission for Time-Triggered flows and high throughput transmission for Best-Effort flows.Cloud computing datacenters have become the technological foundation of contemporary social development,and the datacenter network is the key to the performance of datacenters.Therefore,the research on the optimal scheduling strategy of the datacenter network,has strong practicability and popularization value.