Utilizing multiple access networks such as 5G, 4G, and Wi-Fi simultaneously can lead to increased robustness, resiliency, and capacity for mobile users. However, transparently implementing packet distribution over multiple paths within the core of the network faces multiple challenges including scalability to a large number of customers, low latency, and high-capacity packet processing requirements. In this paper, we offload congestion-aware multipath packet scheduling to a smartNIC. However, such hardware acceleration faces multiple challenges due to programming language and platform limitations. We implement different multipath schedulers in P4 with different complexity in order to cope with dynamically changing path capacities. Using testbed measurements, we show that our CMon scheduler, which monitors path congestion in the data plane and dynamically adjusts scheduling weights for the different paths based on path state information, can process more than 3.5 Mpps packets 25 μs latency.

Utilizing multiple access technologies such as 5G,4G, and Wi-Fi within a coherent framework is currentlystandardized by 3GPP within 5G ATSSS. Indeed, distributingpackets over multiple networks can lead to increased robustness,resiliency and capacity. A key part of such a framework isthe multi-access proxy, which transparently distributes packetsover multiple paths. As the proxy needs to serve thousands ofcustomers, scalability and performance are crucial for operatordeployments. In this paper, we leverage recent advancementsin data plane programming, implement a multi-access proxybased on the MP-DCCP tunneling approach in P4 and hardwareaccelerate it by deploying the pipeline on a smartNIC. Thisis challenging due to the complex scheduling and congestioncontrol operations involved. We present our pipeline and datastructures design for congestion control and packet schedulingstate management. Initial measurements in our testbed showthat packet latency is in the range of 25 μs demonstrating thefeasibility of our approach.

The emergence of Network Functions Virtualization (NFV) is being heralded as an enabler of the recent technologies such as 5G/6G, IoT and heterogeneous networks. Existing NFV monitoring frameworks either do not have the capabilities to express the range of telemetry items needed to perform management or do not scale to large traffic volumes and rates. We present IntOpt, a scalable and expressive telemetry system designed for flexible NFV monitoring using active probing and P4. IntOpt allows us to specify monitoring requirements for individual service chain, which are mapped to telemetry item collection jobs that fetch the required telemetry items from P4 programmable data-plane elements. We propose mixed integer linear program (MILP) as well as a simulated annealing based random greedy (SARG) meta-heuristic approach to minimize the overhead due to active probing and collection of telemetry items. Using P4-FPGA, we benchmark the overhead for telemetry collection. Our numerical evaluation shows that the proposed approach can reduce monitoring overheads by 39% and monitoring delays by 57%. Such optimization may as well enable existing expressive monitoring frameworks to scale for larger real-time networks. 

Industrial Internet of Things (I-IoT) applications require a large number of sensor data to be processed under tight delay and jitter constraints. In such applications, flexible event detection and fast reaction to critical events is an important building block. Traditional approaches use either proprietary networks and dedicated hardware or transmit sensor data towards processing elements in the Cloud or at the Network Edge, using distributed stream processing frameworks. For scalability, a large number of servers are needed and processing on commodity CPUs typically involves high and unpredictable latency. In this paper, we explore how programmable data planes can be used to detect events flexibly and trigger customized and programmable actions directly from the switch program or the programmable network interface card (SmartNIC). We present FastReact-PS, an event-based publish/subscribe I-IoT processing framework in P4 language, which can be flexibly customized from the control plane. Together with stateful processing, FastReact-PS supports windowed time series analysis as well as complex event detection and processing based on boolean logic directly in the data plane of newly emerging programmable networking devices. The logic can be adjusted dynamically from the control plane without the need for recompilation. We implement FastReact-PS in P4 and evaluate it on both a SmartNIC and a DPDK-based software switch running in user space. Our evaluation shows that the latency is reduced by one order of magnitude compared to end-host based approaches at significantly lower jitter while being scalable to processing up to 11 million events per second.

Next generation networks aim to increase network convergence by allowing a single network architecture to serve diverse traffic types ranging from high-bandwidth video streaming to low-latency industrial automation, while meeting their respective service level requirements. Such a converged network architecture puts high requirements on flexibility, interoperability, and resilience. While current networks exhibit some degree of network convergence, they may not reach the level of interoperability required for future application areas. This is particularly prevalent in networks that depend heavily on closed and proprietary equipment, such as industrial automation and small cell backhaul networks. Recently, Software Defined Networking (SDN) and Network Function Virtualization (NFV) have been proposed as solutions for increased network flexibility. By separating and logically centralizing the network control plane, SDN allows for dynamic control of the network infrastructure. NFV, on the other hand, enables flexibility and scalability through the virtualization and orchestration of network functions.

In this thesis, we investigate how SDN and NFV can be used to make next generation networks more reliable, flexible and programmable. We focus mainly on three areas: resiliency, monitoring, and control. First, we look at the usage of SDN to enable in-network resiliency in wireless access, backhaul and industrial automation networks. Next, we design and evaluate FastReact, a switch program that allows industrial automation networks to partially offload their distributed application logic to the data plane, reducing end to end latency and increasing network resiliency. Finally, we propose combining FastReact control with in-network telemetry event detection, significantly increasing the monitoring capacity by selectively discarding redundant telemetry information in the data plane.

Next generation computer networks aim to provide a single network architecture, which can support any type of service, ranging from high-bandwidth video streaming to low-latency industrial automation. Those services have a wide range of network requirements that must be supported by a single converged network, which puts high requirements on flexibility, interoperability, and resilience.

Recently, Software Defined Networking (SDN) and Network Function Virtualization (NFV) have been proposed as solutions for increased network flexibility. By separating and logically centralizing the network control plane, SDN allows for dynamic control of the network infrastructure. NFV, on the other hand, enables flexibility and scalability through the virtualization and orchestration of network functions.

In this thesis, we investigate how SDN and NFV can be used to make next generation networks more reliable, flexible and programmable. We focus mainly on three different areas: resiliency, monitoring, and control, and how they can be improved upon through using SDN. 

Certificate Transparency (CT) requires that every certificate which is issued by a certificate authority must be publicly logged. While a CT log can be untrusted in theory, it relies on the assumption that every client observes and cryptographically verifies the same log. As such, some form of gossip mechanism is needed in practice. Despite CT being adopted by several major browser vendors, no gossip mechanism is widely deployed. We suggest an aggregation-based gossip mechanism that passively observes cryptographic material that CT logs emit in plain text, aggregating at packet processors (such as routers and switches) to periodically verify log consistency off-path. In other words, gossip is provided as-a-service by the network. Our proposal can be implemented for a variety of programmable packet processors at line-speed without aggregation distinguishers (throughput), and, based on 20 days of RIPE Atlas measurements that represent clients from 3500 autonomous systems, we show that significant protection against split-viewing CT logs can be achieved with a realistic threat model and an incremental deployment scenario.

Managing and scaling virtual network function(VNF) service chains require the collection and analysis ofnetwork statistics and states in real time. Existing networkfunction virtualization (NFV) monitoring frameworks either donot have the capabilities to express the range of telemetryitems needed to perform management or do not scale tolarge traffic volumes and rates. We present IntOpt, a scalableand expressive telemetry system designed for flexible VNFservice chain network monitoring using active probing. IntOptallows to specify monitoring requirements for individual servicechain, which are mapped to telemetry item collection jobsthat fetch the required telemetry items from P4 (programmingprotocol-independent packet processors) programmable dataplaneelements. In our approach, the SDN controller creates theminimal number of monitoring flows to monitor the deployedservice chains as per their telemetry demands in the network.We propose a simulated annealing based random greedy metaheuristic(SARG) to minimize the overhead due to activeprobing and collection of telemetry items. Using P4-FPGA, webenchmark the overhead for telemetry collection and compareour simulated annealing based approach with a na¨ıve approachwhile optimally deploying telemetry collection probes. Ournumerical evaluation shows that the proposed approach canreduce the monitoring overhead by 39% and the total delays by57%. Such optimization may as well enable existing expressivemonitoring frameworks to scale for larger real-time networks.

In-Band Network Telemetry (INT) is a novel framework for collecting telemetry items and switch internal state information from the data plane at line rate. With the suppor programmable data planes and programming language P4,switches parse telemetry instruction headers and determine which telemetry items to attach using custom metadata. At the network edge, telemetry information is removed and the original packets are forwarded while telemetry reports are sent to a distributed stream processor for further processing by a network monitoring platform. In order to avoid excessive load on the stream processor, telemetry items should not be sent for each individual packet but rather when certain events are triggered. In this paper, we develop a programmable INT event detection mechanism in P4 that allows customization of which events to report to the monitoring system, on a per-flow basis, from the control plane. At the stream processor, we implement a fast INT report collector using the kernel bypass technique AF XDP, which parses telemetry reports and streams them to a distributed Kafka cluster, which can apply machine learning, visualization and further monitoring tasks. In our evaluation, we use realworld traces from different data center workloads and show that our approach is highly scalable and significantly reduces the network overhead and stream processor load due to effective event pre-filtering inside the switch data plane. While the INT report collector can process around 3 Mpps telemetry reports per core, using event pre-filtering increases the capacity by 10-15x.

Providing network reliability as well as low and predictable latency is important especially for Industrial Automation and Control Networks. However, diagnosing link status from the control plane has high latency and overhead. In addition, the communication with the industrial controller may impose additional network latency. We present FastReact - a system enabling In-Network monitoring, control and caching for Industrial Automation and Control Networks. FastReact outsources simple monitoring and control actions to evolving programmable data planes using the P4 language. As instructed by the Industrial Controller through a Northbound API, the SDN controller composes control actions using Boolean Logic which are then installed in the data plane. The data plane parses and caches sensor values and performs simple calculations on them which are connected to fast control actions that are executed locally. For resiliency, FastReact monitors liveness and response of sensors/actuators and performs a fast local link repair in the data plane if a link failure is detected. Our testbed measurement show that FastReact can reduce the sensor/actuator delay while being resilient against several failure events.

In computer networking, failures, such as breaking equipment, cable cuts, power failures and human errors continuously cause communication interruptions. Such failures may result in dissatisfied customers, loss of product reputation, violation of SLAs and even critical failures in industrial systems. SDN, which logically centralizes the control plane, is an emerging technology in computer networking. The global view provided by the SDN controller can be used to reconfigure the network in case of a link failure. However, this reconfiguration may take too long for high availability networks. With the introduction of proactive link repair, backup paths are preinstalled in the forwarding devices, reducing path recovery time.

This thesis addresses the usage of SDN to provide resiliency in high availability networks. First, we consider how SDN can be used for increasing the reliability of ICNs. Second, we investigate how similar technologies could be applied to deal with fast channel attenuation and resulting outage in mmWave backhaul networks. Finally, we look at CloudMAC-based Wireless LAN, and how SDN-enabled QoS improvements could improve connection reliability.

In computer networking, failures, such as breaking equipment, cable cuts, power failures and human errors continuously cause communication interruptions. Such failures may result in dissatisfied customers, loss of product reputation, violation of Service Level Agreements and even critical failures in industrial systems. Recently, the concept of Software Defined Networking (SDN) was introduced. SDN opens up and centralizes the control plane, which allows designing networks more resilient to failures.

In this thesis, we address the usage of SDN in order to provide resiliency in high availability networks. First, we consider how SDN enabled, proactive failure recovery can be used to provide the reliability required in Industrial Control Networks (ICNs). We also investigate how the same approach could be applied to mmWave backhaul networks to cope with fast channel attenuation and the resulting outage. Through extensive experiments, we can demonstrate an increase in reliability for both ICNs and mmWave backhaul networks. Second, we look at CloudMAC-based Wireless LAN, and how SDN-enabled traffic control algorithms could improve connection reliability. Through our experiments we can show that both discriminatory and non-discriminatory algorithms significantly increase the connection reliability. In combination, these results serve to strengthen the image of SDN as a provider of resilient, high-availability networks.