Congestion Control Mechanisms for Inter-Datacenter Networks
Applications running in geographically distributed setting are becoming prevalent. Large-scale online services often share or replicate their data into multiple data centers (DCs) in different geographic regions. Driven by the data communication need of these applications, inter-datacenter network (IDN) is getting increasingly important.However, we find congestion control for inter-datacenter networks quite challenging. Firstly, the inter-datacenter communication involves both data center networks (DCNs) and wide-area networks (WANs) connecting multiple data centers. Such a network environment presents quite heterogeneous characteristics (e.g., buffer depths, RTTs). Existing congestion control mechanisms consider either DCN or WAN congestion, while not simultaneously capturing the degree of congestion for both.
Secondly, to reduce evolution cost and improve flexibility, large enterprises have been building and deploying their wide-area routers based on shallow-buffered switching chips. However, with legacy congestion control mechanisms (e.g., TCP Cubic), shallow buffer can easily get overwhelmed by large BDP (bandwidth-delay product) wide-area traffic, leading to high packet losses and degraded throughput.
This thesis describes my research efforts on optimizing congestion control mechanisms for the inter-datacenter networks. First, we design GEMINI - a practical congestion control mechanism that simultaneously handles congestions both in DCN andWAN. Second, we present FlashPass - a proactive congestion control mechanism that achieves near zero loss without degrading throughput under the shallow-buffered WAN. Extensive evaluation shows their superior performance over existing congestion control mechanisms.
Dynamic Resource Provisioning of a Scalable E2E Network Slicing Orchestration System
Network slicing allows different applications and network services to be deployed on virtualized resources running on a common underlying physical infrastructure. Developing a scalable system for the orchestration of end-to-end (E2E) mobile network slices requires careful planning and very reliable algorithms. In this paper, we propose a novel E2E Network Slicing Orchestration System (NSOS) and a Dynamic Auto- Scaling Algorithm (DASA) for it. Our NSOS relies strongly on the foundation of a hierarchical architecture that incorporates dedicated entities per domain to manage every segment of the mobile network from the access, to the transport and core network part for a scalable orchestration of federated network slices. The DASA enables the NSOS to autonomously adapt its resources to changes in the demand for slice orchestration requests (SORs) while enforcing a given mean overall time taken by the NSOS to process any SOR. The proposed DASA includes both proactive and reactive resource provisioning techniques). The proposed resource dimensioning heuristic algorithm of the DASA is based on a queuing model for the NSOS, which consists of an open network of G/G/m queues. Finally, we validate the proper operation and evaluate the performance of our DASA solution for the NSOS by means of system-level simulations.