PortLand

Topology-aware scalable resource management in multi-hop dense networks

The current society is becoming increasingly interconnected and hyper-connected. Communication networks are advancing, as well as logistics networks, or even networks for the transportation and distribution of natural resources. One of the key benefits of the evolution of these networks is to bring consumers closer to the source of a resource or service. However, this is not a straightforward task, particularly since networks near final users are usually shaped by heterogeneous nodes, sometimes even in very dense scenarios, which may demand or offer a resource at any given moment. In this paper, we present DEN2NE, a novel algorithm designed for the automatic distribution and reallocation of resources in distributed environments. The algorithm has been implemented with six different criteria in order to adapt it to the specific use case under consideration. The results obtained from DEN2DE are promising, owing to its adaptability and its average execution time, which follows a linear distribution in relation to the topology size.

A port-based forwarding load-balancing scheduling approach for cloud datacenter networks

Today’s datacenter networks (DCNs) scale is rapidly increasing because of the wide deployment of cloud services and the rapid rise of edge computing. The bandwidth consumption and cost of a DCN are growing sharply with the extensions of network size. Thus, how to keep the traffic balanced is a key and challenging issue. However, the traditional load balancing algorithms such as Equal-Cost Multi-Path routing (ECMP) are not suitable for high dynamic traffic in cloud DCNs. In this paper, we propose a port-based forwarding load balancing scheduling (PFLBS) approach for Fat-tree based DCNs with some new features which can overcome the disadvantages of the existing load balancing methods in the following aspects. Firstly, we define a port-based source-routing addressing scheme, which decreases the switch complexity and makes the table-lookup operation unnecessary. Secondly, based on this addressing scheme, we proposed an effective routing mechanism which can obtain multiple available paths for flow scheduling based in Fat-tree. All the path information is saved in servers and each server only needs to maintain its own path information. Thirdly, we propose an efficient algorithm to implement large flows scheduling dynamically in terms of current link utilization ratio. This method is suitable for cloud DCNs and edge computing, which can reduce the complexity of the switches and the power consumption of the whole network. The experiment results indicate that the PFLBS approach has better performance compared with the ECMP, Hedera and MPTCP approaches, which decreases the flow completion time and improves the average throughput significantly. PFLBS is simple and can be implemented with a few signaling overheads.

Improved Compact Routing Schemes for Random Interconnects

Random topology has been an increasingly favorable approach for designing interconnection networks, as it can provide a combination of low latency and incremental network growth that could not be provided by the traditional rigid topologies. However, the common shortest-path routing in a random interconnect poses a scalability problem, for it requires global network info to make routing decisions and so, the routing table size (RTS) can be very large. Therefore, this manuscript would aim to revisit the well-known research area of landmark-based compact routing and to improve the universal routing schemes for the specific case of random interconnects. It would propose new landmark-based compact routing schemes, using 2 heuristic techniques to select landmarks that are evenly spaced, which would reduce the RTS in the well-known Thorup and Zwick's scheme by up to 18% and produce a shorter average path length.

Energy-Aware Virtual Data Center Migration

Recently, the concept of virtual data center (VDC) has attracted significant attention from researchers. VDC is made up of virtual nodes and virtual links with guaranteed bandwidth. It offers elasticity and flexibility, which means VDC can adjust resources dynamically according to different requirements. Existing studies focus on how to design the optimal embedding algorithm to achieve high success rate for the virtual data center request. However, due to the resource of physical data center changes over time, the optimal solution may become sub-optimal. In this paper, we study the problem of virtual data center migration and propose an energy-aware virtual data center migration algorithm, called CA-VDCM-ACO. This novel algorithm leverages the migration technique to further reduce the energy consumption with the success rate for the physical data center guaranteed. The extensive experiments show that our algorithm is very effective to reduce the energy consumption.

Security Solution for ARP Cache Poisoning Attacks in Large Data Centre Networks

The bridge protocol (Address Resolution Protocol) ARP, integrating Ethernet (Layer 2) and IP protocol (Layer 3) plays a vital role in TCP/IP communication since ARP packet is the first packet generated during any TCP/IP communications and they are the first traffic from the host. In the large data center, as the size of the broadcast domain (i.e., number of hosts on the network) increases consequently the broadcast traffic from the communication protocols like ARP also increases. This paper addresses the problem faced by Layer 2 protocols like insecured communication, scalability issues and VM migration issues. The proposed system addresses these issues by introducing two new types of messaging with traditional ARP and also combat the ARP Cache poisoning attacks like host impersonation, MITM, Distributed DoS by making ARP stateful. The components of the proposed methodology first start the process by decoding the packets, updates the invalid entry made by the user with Timestamp feature and messages being introduced. The system has been implemented and compared with various existing solutions.

Joint virtual machine assignment and traffic engineering for green data center networks

The popularization of cloud computing brings emergency concern to the energy consumption in big data centers. Besides the servers, the energy consumed by the network in a data center is also considerable. Existing works for improving the network energy efficiency are mainly focused on traffic engineering, i.e., consolidating flows and switching off unnecessary devices, which fails to comprehensively consider the unique features in data centers. In this paper, we advocate a joint optimization for achieving energy efficiency of data center networks by proposing a unified optimization framework. In this framework, we consider to take advantage of the application characteristics and topology features, and to integrate virtual machine assignment and traffic engineering. Under this framework, we then devise two efficient algorithms, TE VMA and TER, for assigning virtual machines and routing traffic flows respectively. Knowing the ncommunication patterns of the applications, the TE VMA algorithm is purposeful and can generate desirable traffic conditions for the next-step routing optimization. The TER algorithm makes full use of the hierarchical feature of the topology and is conducted on the multipath routing protocol. The performance of the overall framework is confirmed by both theoretical analysis and simulation results, where up to 50% total energy savings can be achieved, 20% more compared with traffic engineering only approaches.

NetPilot

Driven by the soaring demands for always-on and fast-response online services, modern datacenter networks have recently undergone tremendous growth. These networks often rely on commodity hardware to reach immense scale while keeping capital expenses under check. The downside is that commodity devices are prone to failures, raising a formidable challenge for network operators to promptly handle these failures with minimal disruptions to the hosted services.

Recent research efforts have focused on automatic failure localization. Yet, resolving failures still requires significant human interventions, resulting in prolonged failure recovery time. Unlike previous work, NetPilot aims to quickly mitigate rather than resolve failures. NetPilot mitigates failures in much the same way operators do -- by deactivating or restarting suspected offending components. NetPilot circumvents the need for knowing the exact root cause of a failure by taking an intelligent trial-and-error approach. The core of NetPilot is comprised of an Impact Estimator that helps guard against overly disruptive mitigation actions and a failure-specific mitigation planner that minimizes the number of trials. We demonstrate that NetPilot can effectively mitigate several types of critical failures commonly encountered in production datacenter networks.

Surviving failures in bandwidth-constrained datacenters

Datacenter networks have been designed to tolerate failures of network equipment and provide sufficient bandwidth. In practice, however, failures and maintenance of networking and power equipment often make tens to thousands of servers unavailable, and network congestion can increase service latency. Unfortunately, there exists an inherent tradeoff between achieving high fault tolerance and reducing bandwidth usage in network core; spreading servers across fault domains improves fault tolerance, but requires additional bandwidth, while deploying servers together reduces bandwidth usage, but also decreases fault tolerance. We present a detailed analysis of a large-scale Web application and its communication patterns. Based on that, we propose and evaluate a novel optimization framework that achieves both high fault tolerance and significantly reduces bandwidth usage in the network core by exploiting the skewness in the observed communication patterns.

Refactoring network infrastructure to improve manageability

Managing a home network is challenging because the underlying infrastructure is so complex. Existing interfaces either hide or expose the network's underlying complexity, but in both cases, the information that is shown does not necessarily allow a user to complete desired tasks. Recent advances in software defined networking, however, permit a redesign of the underlying network and protocols, potentially allowing designers to move complexity further from the user and, in some cases, eliminating it entirely. In this paper, we explore whether the choices of what to make visible to the user in the design of today's home network infrastructure, performance, and policies make sense. We also examine whether new capabilities for refactoring the network infrastructure - changing the underlying system without compromising existing functionality - should cause us to revisit some of these choices. Our work represents a case study of how co-designing an interface and its underlying infrastructure could ultimately improve interfaces for that infrastructure.

Managing data transfers in computer clusters with orchestra

Cluster computing applications like MapReduce and Dryad transfer massive amounts of data between their computation stages. These transfers can have a significant impact on job performance, accounting for more than 50% of job completion times. Despite this impact, there has been relatively little work on optimizing the performance of these data transfers, with networking researchers traditionally focusing on per-flow traffic management. We address this limitation by proposing a global management architecture and a set of algorithms that (1) improve the transfer times of common communication patterns, such as broadcast and shuffle, and (2) allow scheduling policies at the transfer level, such as prioritizing a transfer over other transfers. Using a prototype implementation, we show that our solution improves broadcast completion times by up to 4.5X compared to the status quo in Hadoop. We also show that transfer-level scheduling can reduce the completion time of high-priority transfers by 1.7X.

Optimizing a virtualized data center

Many data centers extensively use virtual machines (VMs), which provide the flexibility to move workload among physical servers. VMs can be placed to maximize application performance, power efficiency, or even fault tolerance. However, VMs are typically repositioned without considering network topology, congestion, or traffic routes.

In this demo, we show a system, Virtue, which enables the comparison of different algorithms for VM placement and network routing at the scale of an entire data center. Our goal is to understand how placement and routing affect overall application performance by varying the types and mix of workloads, network topologies, and compute resources; these parameters will be available for demo attendees to explore.

Energy proportional datacenter networks

Numerous studies have shown that datacenter computers rarely operate at full utilization, leading to a number of proposals for creating servers that are energy proportional with respect to the computation that they are performing.

In this paper, we show that as servers themselves become more energy proportional, the datacenter network can become a significant fraction (up to 50%) of cluster power. In this paper we propose several ways to design a high-performance datacenter network whose power consumption is more proportional to the amount of traffic it is moving -- that is, we propose energy proportional datacenter networks .

We first show that a flattened butterfly topology itself is inherently more power efficient than the other commonly proposed topology for high-performance datacenter networks. We then exploit the characteristics of modern plesiochronous links to adjust their power and performance envelopes dynamically. Using a network simulator, driven by both synthetic workloads and production datacenter traces, we characterize and understand design tradeoffs, and demonstrate an 85% reduction in power --- which approaches the ideal energy-proportionality of the network.

Our results also demonstrate two challenges for the designers of future network switches: 1) We show that there is a significant power advantage to having independent control of each unidirectional channel comprising a network link, since many traffic patterns show very asymmetric use, and 2) system designers should work to optimize the high-speed channel designs to be more energy efficient by choosing optimal data rate and equalization technology. Given these assumptions, we demonstrate that energy proportional datacenter communication is indeed possible.