Abstract:

Container networking is now an important part of cloud virtualization architectures. It provides network access for containers by connecting both virtual and physical network interfaces. The performance of container networking has multiple dependencies, and each factor may significantly affect the performance. In this paper, we perform systematic experiments to study the performance of container networking. We qualify influencing factors and explain the cause of overhead.

Abstract:

One of the main objectives of any cluster management system is the maximization of cluster resource utilization (CRU). In this paper, we argue that there is a dilemma underlying the challenge of maximizing CRU, as soon as network resources enter the picture. In contrast to local resources which can be handled in a more isolated fashion, global network resources are namely shared, and their allocation is intertwined with that of local resources. For effective resource management, either applications thus have to learn more about the infrastructure, or the resource manager has to understand application semantics – both options violate the separation of applications from the underlying infrastructure strived for by resource managers. This paper makes the case for a resource management system that addresses the dilemma, and presents first ideas.

Abstract:

Network Functions Virtualization (NFV) has enabled operators to dynamically place and allocate resources for network services to match workload requirements. However, unbounded end-to-end (e2e) latency of Service Function Chains (SFCs) resulting from distributed Virtualized Network Function (VNF) deployments can severely degrade performance. In particular, SFC instantiations with inter-data center links can incur high e2e latencies and Service Level Agreement (SLA) violations. These latencies can trigger timeouts and protocol errors with latency-sensitive operations.

Traditional solutions to reduce e2e latency involve physical deployment of service elements in close proximity. These solutions are, however, no longer viable in the NFV era. In this paper, we present our solution that bounds the e2e latency in SFCs and inter-VNF control message exchanges by creating micro-service aggregates based on the affinity between VNFs. Our system, Contain-ed, dynamically creates and manages affinity aggregates using light-weight virtualization technologies like containers, allowing them to be placed in close proximity and hence bounding the e2e latency. We have applied Contain-ed to the Clearwater IP Multimedia System and built a proof-of-concept. Our results demonstrate that, by utilizing application and protocol specific knowledge, affinity aggregates can effectively bound SFC delays and significantly reduce protocol errors and service disruptions.

Abstract:

Reproducible and repeatable provisioning of large-scale distributed systems is laborious. The cost of virtual infrastructure and the provisioning complexity are two of the main concerns. The trade-offs between virtual machines (VMs) and containers, the most popular virtualization technologies, further complicate the problem. Although containers incur little overhead compared to VMs, VMs are required for their certain guarantees such as hardware isolation.

In this paper, we present a mutable container provisioning solution, enabling users to switch infrastructure between VMs and containers seamlessly. Our solution allows for significant infrastructure-cost optimizations.We discuss that immutable containers come short for certain provisioning scenarios. However, mutable containers can incur a large time overhead. To reduce the time overhead, we propose multiple provisioning-time optimizations.We implement our solution in Karamel, an open-sourced reproducible provisioning system. Based on our evaluation results, we discuss the cost-optimization opportunities and the time-optimization challenges of our new model.

Abstract:

An ongoing challenge in network system development is in evaluating the design and implementation of its control plane, without actually deploying it at production scale. Existing approaches based on simulation or emulation have various limitations. The emergence of containers offers a new way of emulating network control plane. In this paper, we design a container-based emulation framework, and introduce its implementation with two concrete use cases: a centralized SDN control plane design in Open Virtual Network (OVN) and a decentralized design in Docker’s *libnetwork*. Through sample scalability studies on these two designs, we demonstrate the effectiveness of the proposed approach.

Abstract:

Mobile Edge Computing (MEC) is an emerging network paradigm that provides cloud and IT services at the point of access of the network. Such proximity to the end user translates into ultra-low latency and high bandwidth, while, at the same time, it alleviates traffic congestion in the network core. Due to the need to run servers on edge nodes (e.g., a an LTE-A macro eNodeB), a key element of MEC architectures is to ensure server portability and low overhead. A possible tool that can be used for this purpose is Docker, a framework that allows easy, fast deployment of Linux containers. This paper addresses the suitability of Docker in MEC scenarios by quantifying the CPU consumed by Docker when running two different containerized services: multiplayer gaming and video streaming. Our tests, run with varying numbers of clients and servers, yield different results for the two case studies: for the gaming service, the overhead logged by Docker increases only with the number of servers; conversely, for the video streaming case, the overhead is not affected by the number of either clients or servers.

Abstract:

FADES is an edge offloading architecture that empowers us to run compact, single purpose tasks at the edge of the network to support a variety of IoT and cloud services. The design principle behind FADES is to efficiently exploit the resources of constrained edge devices through fine-grained computation offloading. FADES takes advantage of MirageOS unikernels to isolate and embed application logic in concise Xen-bootable images. We have implemented FADES and evaluated the system performance under various hardware and network conditions. Our results show that FADES can effectively strike a balance between running complex applications in the cloud and simple operations at the edge. As a solid step to enable fine-grained edge offloading, our experiments also reveal the limitation of existing IoT hardware and virtualization platforms, which shed light on future research to bring unikernel into IoT domain.

Abstract:

In this paper, we describe a new approach for managing service function chains in scenarios where data from Internet of Things (IoT) devices are partially processed at the network edge. Our framework is enabled by two emerging technologies, Software-Defined Networking (SDN) and container based virtualization, which ensure several benefits in terms of flexibility, easy programmability, and versatility. These features are well suitable with the increasingly stringent requirements of IoT applications, and allow a dynamic and automated network service chaining. An extensive performance evaluation, which has been carried out by means of a testbed, seeks to understand how our proposed framework performs in terms of computational overhead, network bandwidth, and energy consumption. By accounting for the constraints of typical IoT gateways, our evaluation tries to shed light on the actual deployability of the framework on low-power nodes.