arXiv subjects

David Bermbach

Publications and source records attributed to David Bermbach.

At least 55 records · Page 3Linked to original sources

arXiv2024

GeoFaaS: An Edge-to-Cloud FaaS Platform

The massive growth of mobile and IoT devices demands geographically distributed computing systems for optimal performance, privacy, and scalability. However, existing edge-to-cloud serverless platforms lack location awareness, resulting in inefficient network usage and increased latency. In this paper, we propose GeoFaaS, a novel edge-to-cloud Function-as-a-Service (FaaS) platform that leverages real-time client location information for transparent request execution on the nearest available FaaS node. If needed, GeoFaaS transparently offloads requests to the cloud when edge resources are overloaded, thus, ensuring consistent execution without user intervention. GeoFaaS has a modular and decentralized architecture: building on the single-node FaaS system tinyFaaS, GeoFaaS works as a stand-alone edge-to-cloud FaaS platform but can also integrate and act as a routing layer for existing FaaS services, e.g., in the cloud. To evaluate our approach, we implemented an open-source proof-of-concept prototype and studied performance and fault-tolerance behavior in experiments.

cs.DC↗

arXiv2024

Lessons Learned from Building Edge Software System Testbeds

Edge computing requires the complex software interaction of geo-distributed, heterogeneous components. The growing research and industry interest in edge computing software systems has necessitated exploring ways of testing and evaluating edge software at scale without relying on physical infrastructure. Beyond simulation, virtual testbeds that emulate edge infrastructure can provide a cost-efficient yet realistic environment to evaluate edge software. In this experience paper, we share lessons learned from building a total of five edge software testbeds. We describe pitfalls in architecture and development as well as experiences from having students use our testbed tooling in distributed systems prototyping classes. While we remain confident that building custom testbed tooling is the right approach for edge computing researchers and practitioners alike, we hope this paper allows others to avoid common mistakes and benefit from our experience.

cs.DC↗

arXiv2024

Application-Centric Benchmarking of Distributed FaaS Platforms using BeFaaS

Due to the popularity of the FaaS programming model, there is now a wide variety of commercial and open-source FaaS systems. Hence, for comparison of different FaaS systems and their configuration options, FaaS application developers rely on FaaS benchmarking frameworks. Existing frameworks, however, tend to evaluate only single isolated aspects, a more holistic application-centric benchmarking framework is still missing. In previous work, we proposed BeFaaS, an extensible application-centric benchmarking framework for FaaS environments that focuses on the evaluation of FaaS platforms through realistic and typical examples of FaaS applications. In this extended paper, we (i) enhance our benchmarking framework with additional features for distributed FaaS setups, (ii) design application benchmarks reflecting typical FaaS use cases, and (iii) use them to run extensive experiments with commercial cloud FaaS platforms (AWS Lambda, Azure Functions, Google Cloud Functions) and the tinyFaaS edge serverless platform. BeFaaS now includes four FaaS application-centric benchmarks, is extensible for additional workload profiles and platforms, and supports federated benchmark runs in which the benchmark application is distributed over multiple FaaS systems while collecting fine-grained measurement results for drill-down analysis. Our experiment results show that (i) network transmission is a major contributor to response latency for function chains, (ii) this effect is exacerbated in hybrid edge-cloud deployments, (iii) the trigger delay between a published event and the start of the triggered function ranges from about 100ms for AWS Lambda to 800ms for Google Cloud Functions, and (iv) Azure Functions shows the best cold start behavior for our workloads.

cs.DC↗

arXiv2024

Are Unikernels Ready for Serverless on the Edge?

Function-as-a-Service (FaaS) is a promising edge computing execution model but requires secure sandboxing mechanisms to isolate workloads from multiple tenants on constrained infrastructure. Although Docker containers are lightweight and popular in open-source FaaS platforms, they are generally considered insufficient for executing untrusted code and providing sandbox isolation. Commercial cloud FaaS platforms thus rely on Linux microVMs or hardened container runtimes, which are secure but come with a higher resource footprint. Unikernels combine application code and limited operating system primitives into a single purpose appliance, reducing the footprint of an application and its sandbox while providing full Linux compatibility. In this paper, we study the suitability of unikernels as an edge FaaS execution environment using the Nanos and OSv unikernel tool chains. We compare performance along several metrics such as cold start overhead and idle footprint against sandboxes such as Firecracker Linux microVMs, Docker containers, and secure gVisor containers. We find that unikernels exhibit desirable cold start performance, yet lag behind Linux microVMs in stability. Nevertheless, we show that unikernels are a promising candidate for further research on Linux-compatible FaaS isolation.

cs.DC↗

arXiv2024

A Hybrid Communication Approach for Metadata Exchange in Geo-Distributed Fog Environments

Metadata exchange is crucial for efficient geo-distributed fog computing. Existing solutions for metadata exchange overlook geo-awareness or lack adequate failure tolerance. We propose HFCS, a novel hybrid communication system that combines hierarchical and peer-to-peer elements, along with edge pools. HFCS utilizes a gossip protocol for dynamic metadata exchange. In simulation, we investigate the impact of node density and edge pool size on HFCS performance. We observe a performance improvement for clustered node distributions, aligning well with real-world scenarios. HFCS outperforms a hierarchical and a P2P approach in task fulfillment at a slight cost to failure detection.

cs.DC↗

arXiv2024

Predicting Temporal Aspects of Movement for Predictive Replication in Fog Environments

To fully exploit the benefits of the fog environment, efficient management of data locality is crucial. Blind or reactive data replication falls short in harnessing the potential of fog computing, necessitating more advanced techniques for predicting where and when clients will connect. While spatial prediction has received considerable attention, temporal prediction remains understudied. Our paper addresses this gap by examining the advantages of incorporating temporal prediction into existing spatial prediction models. We also provide a comprehensive analysis of spatio-temporal prediction models, such as Deep Neural Networks and Markov models, in the context of predictive replication. We propose a novel model using Holt-Winter's Exponential Smoothing for temporal prediction, leveraging sequential and periodical user movement patterns. In a fog network simulation with real user trajectories our model achieves a 15% reduction in excess data with a marginal 1% decrease in data availability.

cs.DC↗

arXiv2023

The Early Microbenchmark Catches the Bug -- Studying Performance Issues Using Micro- and Application Benchmarks

An application's performance regressions can be detected by both application or microbenchmarks. While application benchmarks stress the system under test by sending synthetic but realistic requests which, e.g., simulate real user traffic, microbenchmarks evaluate the performance on a subroutine level by calling the function under test repeatedly. In this paper, we use a testbed microservice application which includes three performance issues to study the detection capabilities of both approaches. In extensive benchmarking experiments, we increase the severity of each performance issue stepwise, run both an application benchmark and the microbenchmark suite, and check at which point each benchmark detects the performance issue. Our results show that microbenchmarks detect all three issues earlier, some even at the lowest severity level. Application benchmarks, however, raised false positive alarms, wrongly detected performance improvements, and detected the performance issues later.

cs.DC↗

arXiv2023

Efficiently Detecting Performance Changes in FaaS Application Releases

The source code of Function as a Service (FaaS) applications is constantly being refined. To detect if a source code change introduces a significant performance regression, the traditional benchmarking approach evaluates both the old and new function version separately using numerous artificial requests. In this paper, we describe a wrapper approach that enables the Randomized Multiple Interleaved Trials (RMIT) benchmark execution methodology in FaaS environments and use bootstrapping percentile intervals to derive more accurate confidence intervals of detected performance changes. We evaluate our approach using two public FaaS providers, an artificial performance issue, and several benchmark configuration parameters. We conclude that RMIT can shrink the width of confidence intervals in the results from 10.65% using the traditional approach to 0.37% using RMIT and thus enables a more fine-grained performance change detection.

cs.DC↗

arXiv2023

Supporting UAVs with Edge Computing: A Review of Opportunities and Challenges

Over the last years, Unmanned Aerial Vehicles (UAVs) have seen significant advancements in sensor capabilities and computational abilities, allowing for efficient autonomous navigation and visual tracking applications. However, the demand for computationally complex tasks has increased faster than advances in battery technology. This opens up possibilities for improvements using edge computing. In edge computing, edge servers can achieve lower latency responses compared to traditional cloud servers through strategic geographic deployments. Furthermore, these servers can maintain superior computational performance compared to UAVs, as they are not limited by battery constraints. Combining these technologies by aiding UAVs with edge servers, research finds measurable improvements in task completion speed, energy efficiency, and reliability across multiple applications and industries. This systematic literature review aims to analyze the current state of research and collect, select, and extract the key areas where UAV activities can be supported and improved through edge computing.

cs.DC↗

arXiv2023

Enoki: Stateful Distributed FaaS from Edge to Cloud

Function-as-a-Service (FaaS) is a promising paradigm for applications distributed across the edge-cloud continuum. FaaS functions are stateless by nature, leading to high elasticity and transparent invocation. Supporting stateful applications, however, requires integrating data storage in FaaS, which is not trivial in an edge-cloud environment. We propose Enoki, an architecture for stateful FaaS computing replicated across the edge-cloud continuum. Enoki integrates a replicated key-value store with single-node FaaS systems at edge and cloud nodes in order to provide low-latency local data access for functions without breaking the abstraction of the FaaS programming model. We evaluate Enoki with microbenchmarks on an open-source prototype and demonstrate building a stateful FaaS application with multiple functions distributed over edge and cloud.

cs.DC↗

arXiv2023

Can Orbital Servers Provide Mars-Wide Edge Computing?

Human landing, exploration and settlement on Mars will require local compute resources at the Mars edge. Landing such resources on Mars is an expensive endeavor. Instead, in this paper we lay out how concepts from low-Earth orbit edge computing may be applied to Mars edge computing. This could lower launching costs of compute resources for Mars while also providing Mars-wide networking and compute coverage. We propose a possible Mars compute constellation, discuss applications, analyze feasibility, and raise research questions for future work.

cs.DC↗

arXiv2023

Edge Computing in Low-Earth Orbit -- What Could Possibly Go Wrong?

Large low-Earth orbit (LEO) satellite networks are being built to provide low-latency broadband Internet access to a global subscriber base. In addition to network transmissions, researchers have proposed embedding compute resources in satellites to support LEO edge computing. To make software systems ready for the LEO edge, they need to be adapted for its unique execution environment, e.g., to support handovers in face of satellite mobility. So far, research around LEO edge software systems has focused on the predictable behavior of satellite networks, such as orbital movements. Additionally, we must also consider failure patterns, e.g., effects of radiation on compute hardware in space. In this paper, we present a taxonomy of failures that may occur in LEO edge computing and how they could affect software systems. From there, we derive considerations for LEO edge software systems and lay out avenues for future work.

cs.DC↗

arXiv2023

A Crowdsensing Approach for Deriving Surface Quality of Cycling Infrastructure

Cities worldwide are trying to increase the modal share of bicycle traffic to address traffic and carbon emission problems. Aside from safety, a key factor for this is the cycling comfort, including the surface quality of cycle paths. In this paper, we propose a novel edge-based crowdsensing method for analyzing the surface quality of bicycle paths using smartphone sensor data: Cyclists record their rides which after preprocessed on their phones before being uploaded to a private cloud backend. There, additional analysis modules aggregate data from all available rides to derive surface quality information which can then used for surface quality-aware routing and planning of infrastructure maintenance.

cs.DC↗

arXiv2023

Towards a Benchmark for Fog Data Processing

Fog data processing systems provide key abstractions to manage data and event processing in the geo-distributed and heterogeneous fog environment. The lack of standardized benchmarks for such systems, however, hinders their development and deployment, as different approaches cannot be compared quantitatively. Existing cloud data benchmarks are inadequate for fog computing, as their focus on workload specification ignores the tight integration of application and infrastructure inherent in fog computing. In this paper, we outline an approach to a fog-native data processing benchmark that combines workload specifications with infrastructure specifications. This holistic approach allows researchers and engineers to quantify how a software approach performs for a given workload on given infrastructure. Further, by basing our benchmark in a realistic IoT sensor network scenario, we can combine paradigms such as low-latency event processing, machine learning inference, and offline data analytics, and analyze the performance impact of their interplay in a fog data processing system.

cs.DC↗

arXiv2023

Fog System Configuration Management with CRDTs

Current fog systems rely on centralized and strongly consistent services for configuration management originally designed for cloud systems. In the geo-distributed fog, such systems can exhibit high communication latency or become unavailable in case of network partition. In this paper, we examine the drawbacks of strong consistency for fog configuration management and propose an alternative based on CRDTs. We prototypically implement our approach for the FReD fog data management platform. Early results show reductions of server response times of up to 50\%.

cs.DC↗

arXiv2023

On the Applicability of Network Coordinate Systems for Fog Computing

Identifying the closest fog node is crucial for mobile clients to benefit from fog computing. Relying on geographical location alone us insufficient for this as it ignores real observed client access latency. In this paper, we analyze the performance of the Meridian and Vivaldi network coordinate systems in identifying nearest fog nodes. To that end, we simulate a dense fog environment with mobile clients. We find that while network coordinate systems really find fog nodes in close network proximity, a purely latency-oriented identification approach ignores the larger problem of balancing load across fog nodes.

cs.DC↗

arXiv2023

Managing Data Replication and Distribution in the Fog with FReD

The heterogeneous, geographically distributed infrastructure of fog computing poses challenges in data replication, data distribution, and data mobility for fog applications. Fog computing is still missing the necessary abstractions to manage application data, and fog application developers need to re-implement data management for every new piece of software. Proposed solutions are limited to certain application domains, such as the IoT, are not flexible in regard to network topology, or do not provide the means for applications to control the movement of their data. In this paper, we present FReD, a data replication middleware for the fog. FReD serves as a building block for configurable fog data distribution and enables low-latency, high-bandwidth, and privacy-sensitive applications. FReD is a common data access interface across heterogeneous infrastructure and network topologies, provides transparent and controllable data distribution, and can be integrated with applications from different domains. To evaluate our approach, we present a prototype implementation of FReD and show the benefits of developing with FReD using three case studies of fog computing applications.

cs.DC↗

arXiv2023

The Night Shift: Understanding Performance Variability of Cloud Serverless Platforms

Function-as-a-Service is a popular cloud programming model that supports developers by abstracting away most operational concerns with automatic deployment and scaling of applications. Due to the high level of abstraction, developers rely on the cloud platform to offer a consistent service level, as decreased performance leads to higher latency and higher cost given the pay-per-use model. In this paper, we measure performance variability of Google Cloud Functions over multiple months. Our results show that diurnal patterns can lead to performance differences of up to 15%, and that the frequency of unexpected cold starts increases threefold during the start of the week. This behavior can negatively impact researchers that conduct performance studies on cloud platforms and practitioners that run cloud applications.

cs.DC↗