Doctor of Philosophy (PhD)
Division of Computer Science and Engineering
A data center (DC) has been a fundamental infrastructure for academia and industry for many years. Applications in DC have diverse requirements on communication. There are huge demands on data center network (DCN) control frameworks (CFs) for coordinating communication traffic. Simultaneously satisfying all demands is difficult and inefficient using existing traditional network devices and protocols. Recently, the agile software-defined Networking (SDN) is introduced to DCN for speeding up the development of the DCNCF. Application-awareness preserves the application semantics including the collective goals of communications. Previous works have illustrated that application-aware DCNCFs can much more efficiently allocate network resources by explicitly considering applications needs.
A transfer application task level application-aware software-defined DCNCF (SDDCNCF) for OpenFlow software-defined DCN (SDDCN) for big data exchange is designed. The SDDCNCF achieves application-aware load balancing, short average transfer application task completion time, and high link utilization. The SDDCNCF is immediately deployable on SDDCN which consists of OpenFlow 1.3 switches. The Big Data Research Integration with Cyberinfrastructure for LSU (BIC-LSU) project adopts the SDDCNCF to construct a 40Gb/s high-speed storage area network to efficiently transfer big data for accelerating big data related researches at Louisiana State University.
On the basis of the success of BIC-LSU, a coflow level application-aware SD- DCNCF for OpenFlow-based storage area networks, MinCOF, is designed. MinCOF incorporates all desirable features of existing coflow scheduling and routing frame- works and requires minimal changes on hosts.
To avoid the architectural limitation of the OpenFlow SDN implementation, a coflow level application-aware SDDCNCF using fast packet processing library, Coflourish, is designed. Coflourish exploits congestion feedback assistances from SDN switches in the DCN to schedule coflows and can smoothly co-exist with arbitrary applications in a shared DCN. Coflourish is implemented using the fast packet processing library on an SDN switch, Open vSwitch with DPDK. Simulation and experiment results indicate that Coflourish effectively shortens average application completion time.
Chiu, Chui-Hui, "A Study of Application-awareness in Software-defined Data Center Networks" (2017). LSU Doctoral Dissertations. 4138.
Available for download on Saturday, December 15, 2018