The predominantly content-oriented use of today's networks has led to the development of information-Centric Networking (ICN) paradigms and network architectural designs. At the core of such designs is naming, which organizes and guides the delivery and dissemination of information in the network. This dissertation studies the correctness of such name-based networking, improving it for better scalability, and enhancing its functionality in real-world applications.

As a first topic, we propose Name Space Analysis (NSA), a network verification framework to model and analyze name-based data planes of Named Data Networks. NSA supports primitives fundamentally different from those of traditional host-centric IP network verification, such as checking host-to-content reachability in addition to the traditional host-to-host reachability considerations. We also design and formally analyze name-based delivery for inter-operation of ICNs with existing IP networks.

State-of-the-art ICN designs rely on strictly-hierarchical naming frameworks and pull-based request/response models, which can be inadequate when it comes to complex information structures and many-senders-to-many-receivers delivery. We propose POISE, a name-based and recipient-based publish/subscribe architecture for efficient information dissemination that supports complex graph-based namespaces with a workload-driven namespace graph partitioning for multicast core migration. We demonstrate that POISE achieves better efficiency in timely delivery and elimination of traffic concentration than existing alternatives.

Next, we consider infrastructure-less intermittently-connected network environments, where consistency of replicated databases is especially challenging. We propose CoNICE, a framework to ensure consistent dissemination of updates in such environments. Our proposed name-based coordinator-less consensus method improves completeness and the convergence latency in ordering updates, all with lower communication overhead. We study its correctness and scalability and also demonstrate its performance benefits.

We then look at how names can be learned from the content itself. We propose a framework that integrates Natural Language Processing techniques with Information-Centric dissemination, considering the role of social media-based incident reporting in disaster response. The framework introduces a social media engine to intelligently map social media posts to the right names, to steer posts (content) to the most relevant first responders in a timely manner. We further show how these social media engines can be enhanced, using active and federated learning, for better accuracy.

Our work seeks to improve and adapt computing systems and machine learning (ML) algorithms to match each other’s requirements and capabilities. Due to the high computational demand for Deep Neural Networks (DNNs), accelerators such as GPUs are necessary to achieve low-latency inference. With GPUs getting more powerful, DNNs often fail to utilize a GPU’s parallelism fully. Understanding the GPU resources a DNN model can effectively use while still fulfilling the SLO of users allows us to run multiple DNN models concurrently by spatially sharing the GPU. Our DNN inference framework virtualizes the GPU, adapts to the DNN model, and improves the CPU-GPU coordination, to achieve a much higher aggregate inference throughput compared to other multiplexing techniques. While spatial sharing utilizes the GPU’s resources better, its static resource allocation is unsuitable for dynamic workloads. We propose a spatio-temporal scheduler that provides the right GPU resources for multiple DNNs and meets the inference tasks’ SLOs. Our spatio-temporal scheduler can run more models concurrently and achieve 4 × higher throughput than other scheduling techniques.

Autotuning a DNN model customizes it to match the system’s capabilities. However, current autotuning frameworks do not consider a DNN model’s GPU demand. They produce a sub-optimally tuned model that does not provide the lowest possible latency when inferring with a smaller amount of GPU resources. Our enhanced autotuning produces a tuned model resilient to different amounts of GPU resources available at inference by targeting the appropriate GPU resources during tuning. Our framework enables concurrent autotuning of multiple models by using spatial multiplexing of the GPU, and eliminates several system overheads, thus decreasing tuning time by more than 70%.

We apply our understanding of GPU to the task of localization in multi-user augmented reality (AR) applications. Our multi-user AR framework efficiently offloads AR computation to an edge server and lowers the localization time by 50%. Our implementation utilizes shared memory on the edge server to reduce the time for ’map merging’ between different users’ maps, a task that needs to be performed frequently with multi-user AR. Our approach cuts the map merging time by more than 80% compared to potential multi-user AR approaches.

Our overall approach is to adapt computing resources to the algorithms that use them. It particularly benefits current ML algorithms and other applications that use these accelerators, such as AR. Our techniques can also improve the throughput of the newer generations of accelerators, which offer a significant speedup by using parallel compute engines.

Cloud computing has fundamentally transformed the landscape of computing by leveraging the widespread accessibility of network connectivity. In this space, serverless computing platforms have emerged as key facilitators, promising cost-effective cloud computing capabilities for users. However, existing serverless platforms face several challenges, particularly in the data plane they use for communication between serverless functions. These data plane mechanisms conflict with the event-driven philosophy of serverless computing. They do not help meet the desired low-latency requirements crucial for a variety of services using the cloud. For example, 5G cellular networks are transitioning to a cloud-native design, adopting a Service-based Architecture, to simplify the development and deployment of 5G systems in the cloud. However, this comes with trade-offs, as it may limit performance and hinder the execution of low-latency operations in 5G cellular core network (5GC).

Our first contribution focuses on assessing the overhead of container network interface, the essential networking component in the serverless data plane. Our investigation revealed that current serverless platforms treat inter-function networking as though each function were isolated by a network link in the kernel, a methodology fraught with overhead. We recognize the imperative of facilitating seamless communication within nodes via shared memory processing and design SPRIGHT. SPRIGHT is the first serverless framework that leverages the extended Berkeley Packet Filter (eBPF), an in-kernel event-driven networking subsystem, to supplant long-running stateful components typically used in current serverless platforms. We also optimize the placement engine in the serverless control plane to account for heterogeneity and fairness across competing functions, ensuring overall resource efficiency, and minimizing resource fragmentation.

We then show how an improved serverless design (called LIFL) can be suitable to support fast and efficient model aggregation in Federated Learning (FL), which typically involves varying numbers of heterogeneous clients locally training machine learning models. LIFL adopts the streamlined data plane from SPRIGHT. We further introduce locality-aware placement in LIFL to maximize the benefits of shared memory processing. LIFL precisely scales and carefully reuses server resources for hierarchical aggregation to achieve the highest degree of parallelism while minimizing aggregation time and resource consumption. Our open-source implementation of LIFL achieves significant improvement in resource efficiency and aggregation speed for supporting FL at scale, compared to existing serverful as well as serverless FL systems. Our last contribution introduces L25GC+, a 3GPP-compliant 5G Core designed to support low-latency control plane operations and high-performance user plane packet transmissions. In the control plane, L25GC+ leverages the 3GPP-specified Service-Based Interface (SBI) for exchanging control events between 5G functions. L25GC+ addresses this by redesigning the SBI using shared memory processing instead of kernel networking. We augment asynchronous shared memory I/O with synchronous data exchange primitives and stateful processing to differentiate across user sessions. The optimized design seamlessly integrates with the 3GPP-compliant SBI, demonstrating improved latency, scalability, and user experience on commercial 5G testbeds. In the 5GC data plane, packet classification becomes expensive with a growing number of user sessions. We devise faster packet classification approaches in the 5GC data plane, ensuring high data plane performance, even with an increasing number of user sessions.

Pre-2018 CSE ID: CS2010-0956

Pre-2018 CSE ID: CS2008-0919