Modern Software as a Service (SaaS) platforms, such as the Microsoft Power Platform, simplify maintenance efforts through abstraction of underlying layers such as managing server infrastructure or OS-level configurations. Nevertheless, they still allow custom code extensions to extend and customize the application behavior. Yet, this implies restricted control and limited insights with respect to performance.Often, built-in performance reportings do not provide enough detail for efficient problem root cause analysis. Open standards, such as OpenTelemetry, make it easier to implement a more advanced aligned tracing ecosystem. Yet, while monitoring on traditional, non-cloud systems has been thoroughly studied, collecting telemetry data on SaaS systems faces many more complex challenges such as sandboxing, e.g., limited API access, or collecting end-to-end traces through multiple distributed entry points. Besides platform experience, deep technical knowledge about serialization, transmission, and caching costs is essential to understand the process and assess the performance overhead introduced by collecting telemetry data.In this work, we discuss common challenges regarding SaaS observability and the performance impact that telemetry data collection introduces. For this, we modify Intermediate Language (IL) code, i.e., already compiled code, on the Microsoft Power Platform to collect telemetry data for each method execution. A key finding is that a significant and often unreported portion of performance overhead originates from the SaaS platform itself, a crucial distinction from the overhead introduced by telemetry collection. Finally, we propose a concept for adaptive telemetry collection based on execution plan prediction, which aims to balance observability detail with performance overhead by estimating monitoring costs.
Cloud-native deployments are typically ad-hoc and fragmented, leading to repeated misconfigurations and increased security risk. A challenge that arises often in modern deployment processes is the use of diverse heterogeneous technologies, such as Infrastructure-as-Code frameworks and cloud-specific tools. This causes a technological lock-in, making migration, adoption of alternative solutions, and supporting multi-cloud environments difficult. Another issue that further complicates current deployment processes is the collaboration of multiple actors during the process, such as developers, cloud architects, and business decision-makers whose roles are not well defined.To address these challenges, we advocate for a constraint-based approach of deployment. We emphasize security and compliance where constraints are declarative and treated as first-class inputs. This process would enable the specification of high-level constraints (e.g. security, cost, legal, etc.) and technical requirements in a unified and structured manner. This makes deployments portable across environments and cloud providers, auditable, and resilient to evolution. By introducing clear role separation and constraints validation, our approach promotes accountability while reducing the cognitive load on individual deployment actors.We propose a framework that enables actors involved in the deployment to specify constraints at various stages (architecture level, implementation level, etc.). These constraints are evaluated and verified by a constraint solver, and results into a set of feasible deployment options. Deployment scripts are generated from one of the feasible deployments in order to deploy the application. It is worth stressing that our intention is not to replace existing deployment workflows, but rather to provide a complementary approach.
Serverless computing simplifies application deployment by removing the need for infrastructure management, with REST APIs being the common interface. Data persistence is essential for serverless applications because serverless functions are stateless and short-lived. To retain information across function executions, relational databases are commonly used to persist data, to ensure continuity, scalability, and reliable management. However, REST can lead to inefficiencies such as data over-fetching and under-fetching, which impact performance. GraphQL is a data query and manipulation language that supports client queries that specify what data is to be retrieved or modified. GraphQL resolvers fetch and transform data as specified by client queries.This paper compares the performance and scalability of GraphQL APIs as a database interface middleware in contrast to REST APIs implemented using serverless functions. We compare a GraphQL API implemented using the managed GraphQL AWS AppSync service, an unmanaged GraphQL API hosted using Apollo Server, and a traditional REST API implemented via the Amazon API Gateway and AWS Lambda. We compare these alternatives using clients implemented with serverless AWS Lambda functions, and also a local machine, an Amazon virtual machine (VM), and a Google VM. Our data APIs accessed a managed Amazon Aurora PostgreSQL cluster populated with the U.S. Centers for Medicare & Medicaid Services (CMS) Open Payments dataset. Our GraphQL implementations show content-dependent performance compared to REST, with Apollo Server demonstrating 25-67% faster average round-trip times vs. REST for most operations, but worse scalability than REST with very high concurrent workloads. Our findings provide practical guidance for developers selecting serverless architectures for GraphQL and REST APIs based on specific application requirements, network conditions, and expected request volumes.
CPILS is a sustainable education system consisting of a private cloud core and peripherals for the interaction with users and the environment. It is aimed at upskilling students on societally relevant topics as well as technology to overcome uncertainties and achieve autarky. Currently, two instances of CPILS are deployed in universities as immersive learning spaces for mixed-discipline students. In this paper, we present the adaptive software design and technology decisions behind CPILS and demonstrate the resulting implementation.
This study investigates whether cloud provider solutions offer cost advantages over private cloud infrastructures in large-scale government settings. Focusing on the case of Greek Governmental Cloud (GCloud), we conduct a comprehensive cost comparison across three deployment models: the existing on-premise infrastructure, an IaaS cloud provider solution and a PaaS cloud provider solution. The cost analysis considers a five-year operational time window and uses real procurement data and updated discount policies. We take into consideration that the organization already has data center infrastructure and explores the migration options to cloud provider solutions. The cost analysis findings show that although provider solution can match or slightly undercut on-premise costs under specific conditions where agility is important, there is no financial advantage regarding standard operation needs. This real-world case study of the Greek Government Cloud highlights that cloud migration decisions for large organizations, with existing IT infrastructures, should be taken after thorough cost evaluation, accurate workload analysis, and a deep understanding of dynamic pricing models. Moreover, this study offers practical recommendations to decision-makers that navigate complex infrastructure choices. While infrastructure choices involve factors such as performance, security, and operational flexibility, this study intentionally focuses solely on the cost, aiming to provide a clear and quantifiable economic comparison. In general, this study aims to open a discussion about whether Cloud solutions are a panacea, especially in government settings.
Today’s cloud native applications are often built using a service-oriented architecture supported by many microservices hosted using serverless Function-as-a-Service (FaaS) platforms. The vast majority of serverless function benchmarks and tests are performed using a single-client machine to generate workloads against highly scalable serverless backends. However, single-client test engines can lack sufficient computational resources and network bandwidth to adequately stress serverless backends. Testing tools such as Apache JMeter support orchestrating tests using multiple client nodes, but lack the ability to orchestrate sophisticated tests with distributed workload patterns common with real-world serverless workloads. In this paper, we introduce Distributed FaaSRunner, a distributed test tool which supports the ability to reproduce multi-node serverless workloads using traces derived by ingesting serverless function log files and randomly generated workload traces. We test Distributed FaaSRunner’s ability to precisely reproduce serverless function request dispatch and arrival time using various test cluster configurations. Using globally distributed clients, we predict latency to adjust workload trace event dispatch times to reproduce original request arrival latency. We demonstrate that Distributed FaaSRunner can reproduce both temporal and spatial characteristics of serverless workloads, enabling new capabilities to assess performance of FaaS platforms beyond traditional load testing.
We demonstrate a new customizable toolchain for validating Kubernetes manifests. The toolchain supports multiple types of validation, including validations against JSON schema, Kubernetes built-in rules, best practices, and custom business rules expressed in Common Expression Language (CEL). It also enables the detection of inconsistencies within a defined set of rules. The main goal is to highlight the value of each validation type and demonstrate how their combination improves the robustness of Kubernetes manifests, ensuring reliable deployments.
This paper addresses the challenges of hosting multiple GPU workloads in resource-constrained edge environments where GPU computation demands often exceeds availability. Traditional approaches to GPU sharing, such as NVIDIA MPS and MIG, have limitations regarding dynamic priority adjustments and hardware compatibility, as several of them require specialized data-center oriented GPUs. To overcome these limitations, we propose a novel, hardware-agnostic GPU spatio-temporal sharing and scheduling mechanism for GPU-intensive, long-running jobs in shared edge clusters requiring non-to-minimum changes in the workload source code. A Local GPU Scheduler Component, can prioritize the execution of each workload based on a set of scheduling policies, including Priority Based Scheduling and Real-Time, Laxity based scheduling in order to achieve the SLOs of time-critical real-time workloads that runs concurrently in the shared GPU. Extensive evaluation on a set of non specialized GPUs with different architectures provides fully insight on the efficiency and the efficacy of the proposed technique.
Cloud computing has become integral to modern technology infrastructure, supporting a wide range of services from e-commerce to AI applications. Chameleon is a large-scale, configurable testbed designed to enable edge-to-cloud research through full bare-metal provisioning, virtualization, and diverse hardware resources, which is built on a leading open source cloud platform OpenStack. However, monitoring Chameleon’s heterogeneous infrastructure is challenging, particularly across Open-Stack services and hardware components. Traditional threshold-based alerting methods struggle to keep up with the scale and complexity of such environments. In this work, we present an anomaly detection framework for OpenStack services in the Chameleon Cloud. We curate and publish the first dataset of resource usage metrics collected from OpenStack control plane services. We evaluate four state-of-the-art unsupervised multivariate time series models, namely TranAD, Prodigy, USAD, and OmniAnomaly, on this dataset and share key insights from deploying them. Our findings indicate that for our use case, while all models achieve high F1 scores, training with three days of healthy data effectively balances training cost and detection accuracy.
Edge computing enables time-sensitive applications by bringing computation closer to data sources. However, existing orchestration frameworks such as Kubernetes often overlook performance variability, leading to inefficient resource management. While recent advances have introduced performance-aware scheduling, load balancing remains largely reactive. In particular, performance-aware load balancers are still missing from mainstream systems, limiting their ability to meet subsecond latency requirements. This paper presents a unified extension to orchestration frameworks that integrates both performance-aware scheduling and load balancing. We design lightweight performance predictors, trained on historical monitoring data, to estimate performance fluctuations. Variability predictors inform the scheduler, while Round-Trip Time (RTT) predictors guide load balancing decisions. Our approach is validated using a Single Particle Analysis (SPA) application deployed on a Kubernetes-based edge computing testbed. Results demonstrate that the predictors achieve approximately 85% accuracy in capturing RTT variability. Simulation results show reductions in application completion times ranging from 50% to 78% for scheduling and 68% for load balancing, highlighting the effectiveness of predictive orchestration in dynamic edge/cloud environments.
This paper presents REST-Q, a framework to measure the energy consumption in multi-tier web applications. Previous studies have shown that technology choices and configurations significantly affect the energy consumption of these systems. Notably, the complexity and variability of web frameworks make it difficult to capture the impact of design choices on the energy consumption of the deployed systems. REST-Q delivers a reproducible infrastructure to evaluate such impacts.
Performance regressions in large-scale software systems can lead to substantial resource inefficiencies, making their early detection critical. Frequent benchmarking is essential for identifying these regressions and maintaining service-level agreements (SLAs). Performance benchmarks, however, are resource-intensive and time-consuming, which is a major challenge for integration into Continuous Integration / Continuous Deployment (CI/CD) pipelines. Although numerous benchmark optimization techniques have been proposed to accelerate benchmark execution, there is currently no practical system that integrates these optimizations seamlessly into real-world CI/CD pipelines. In this vision paper, we argue that the field of benchmark optimization remains under-explored in key areas that hinder its broader adoption. We identify three central challenges to enabling frequent and efficient benchmarking: (a) the composability of benchmark optimization strategies, (b) automated evaluation of benchmarking results, and (c) the usability and complexity of applying these strategies as part of CI/CD systems in practice. We also introduce a conceptual cloud-based benchmarking framework handling these challenges transparently. By presenting these open problems, we aim to stimulate research toward making performance regression detection in CI/CD systems more practical and effective.
Distributed computer vision deployment frameworks aim to reduce end-to-end latency and backbone traffic by distributing computation across edge and cloud tiers. In typical architectures, nodes can process a portion of incoming image streams locally or in an edge server and offload the remainder to the cloud. However, runtime decisions about task scheduling and offloading often impose a significant computational burden on nodes, leading to unnecessary runtime overhead and underutilization of hardware resources for deployed applications. To address these challenges, we propose an offline mixed-integer optimisation framework that partitions workloads into local edge processing and deterministic offloading to the cloud tier. Each edge node handles multiple image streams by deploying parallel processing pipelines. Our framework enforces quality-of-service (QoS) constraints while maximising the proportion of computation performed at the edge. By precomputing an optimal configuration, the edge node can meet latency and resource requirements without incurring the overhead of on-the-fly optimisation. We evaluate our approach against several standard image stream processing approaches and demonstrate that our method satisfies QoS targets with lower processing delay and higher resource utilisation. We demonstrate that as the number of sensors increases, our framework can adjust the deployed configuration to enhance on-edge processing.
Cloud-centric services are becoming the norm in modern IT and Operational Technology (OT) platforms, where cybersecurity risks are also on the rise. The evolving regulatory landscape within Europe, exemplified by the Network and Information Security 2 (NIS2) directive and the Cyber Resilience Act (CRA), further amplifies the necessity for rigorous compliance measures. The presumption of conformity for platforms certified under EU-recognized certification schemes, as outlined by the CRA, is anticipated to promote the certification of IT and OT platforms. Nevertheless, the certification process for these platforms is challenging due to the complexity of cloud architectures and the constantly evolving threats, which require continuous adaptation. Furthermore, both NIS2 and CRA introduce new mandates, including the obligation to manage risks, report incidents to relevant authorities, inform customers about vulnerabilities, and provide relevant mitigation strategies. Consequently, there exists an urgent demand for tools and frameworks that support sustained certification in the cloudification of IT/OT platforms. This paper introduces the Continuous Certification Readiness Framework (CCRF), which is engineered to automate tasks related to certification preparation and support ongoing compliance assessments, thereby enabling organizations to effectively manage risks and uphold a high level of assurance within their cloud environments.
Multi-access Edge Computing (MEC) in Edge-cloud computing continuum environments is an emerging technology that enables locality-aware low-latency processing for requests from edge devices. Confidential Containers (CoCo) are an emerging confidential computing technology designed for protecting in-use data in cloud datacenters. Although containers in a MEC layer can help reduce latency for requests from edge devices, the current increasing demands for security features such as confidential computing and memory integrity protection will seriously affect expected end-to-end latencies of microservice applications. In this paper, we build a CoCo environment using Kata containers on the MEC layer and attempt to evaluate microservice-level application behavior for the advanced security features. Especially, we focus on breaking down the latency into memory access level, microservice component level, and the end-to-end latency level. From the results, we observe that 9% overhead is incurred for each memory access when applying confidential computing with memory integrity protection. At the individual microservices level, we observed the latency for gRPC communication and memcached access are increased up to 40 %. We also reveal that the latency overheads in microservice-level and application-level are from twice to 4 times larger than those observed in memory access level.
The ongoing shift from Cloud Computing to Edge Computing paradigm brings geo-distributed data management challenges back to the forefront, particularly in the context of large-scale data processing.This article explores data replication, a key strategy for improving system performance in geo-distributed computing environments. While replication can improve efficiency, naive strategies often result in excessive and unnecessary data transfers, leading to inefficient resource utilization, particularly in infrastructures interconnected via heterogeneous network links. To address these limitations, we propose a replication strategy that estimates the utility of each potential replica before deployment. The approach is first evaluated in homogeneous environments and then extended to heterogeneous settings with varying network characteristics. Simulation results show that our method significantly reduces data transfers while maintaining high execution efficiency, achieving a balanced trade-off between performance and resource consumption.
In-memory key-value stores are critical caching infrastructure for numerous cloud services. Unlike traditional CPU caches that often assume uniform item sizes, cloud caches frequently handle heterogeneous-sized objects, introducing significant challenges in cache management, particularly in shared multi-tenant environments. Existing cache sharing solutions designed for uniform-sized objects are often not optimized for these scenarios.This paper presents a probabilistic cache sharing scheme that dynamically adapts eviction probabilities across different traffic classes based on their performance. Our approach redistributes cache space at eviction events, reclaiming space probabilistically from one class to serve the needs of another. Our scheme operates independently of the underlying per-class eviction policies and adapts to time-varying traffic patterns. We evaluate our approach using real-world traces with heterogeneous object sizes, demonstrating its effectiveness in dynamically allocating cache space and improving overall cache performance in shared environments.
Ten years after the drafting of the Paris Agreement, the objective to reduce greenhouse gas emissions by half by 2030 seems difficult to reach. The impact of ICT technologies is growing year after year. Data centers have a large impact in this domain. As a lot of research is focused on improving energy efficiency of data centers, one could ask if this improvement would suffice to reduce greenhouse gas emissions. To study the impact of the energy efficiency on the long term, we propose to model a data center that has a limited power capacity and is regularly refreshed with new servers, more energy-efficient than the previous ones, to cope with its workload. Our results explore various growth rates for the load, and the energy efficiency and show that without reduction of usage of data center resources, the improvements in energy efficiency will not be not enough to reach the Paris Agreement’s objective.
This demo showcases DistWalk, an open-source distributed workload emulator designed to study the end-to-end latency implications of Linux-based systems. DistWalk is capable of deploying sequences of compute, network, and storage operations arranged within graph-like topologies, to be carried out across multiple servers. It supports a variety of communication protocols and traffic patterns, and enables the customization of several factors, such as the duration and parallelism of compute-intensive operations, the network security and connection handling strategy, and the I/O data access and synchronization mode, among others. DistWalk can be used to experiment with a variety of deployment models for distributed workloads, from bare-metal to virtualized or containerized environments, e.g., using Cloud/Edge infrastructures, OpenStack, Kubernetes, or other orchestrators. This allows Cloud/Edge researchers and developers to perform experimental comparisons of the latency achievable by distributed workload patterns across a wide range of system-level configurations.
Our prior work on single-metric near real-time anomaly detection is extended in this paper through the generalization of a model that was initially developed for the monitoring of CPU utilization anomalies in Vodafone’s Network Functions Virtualization (NFV) infrastructure. The initial generalization involves the model being adapted to other critical infrastructure KPIs, with a specific focus placed on average Network and Memory usage. Subsequently, a significant reduction in the model’s free parameters is introduced, with the original count of 13 being decreased to a single parameter. Building upon this refined single-metric model, a novel multi-metric anomaly detection model is then constructed. The quality of anomaly detection is demonstrably enhanced by this model through a substantial reduction in the incidence of both false positive and false negative classifications. Empirical results from an experiment conducted on real-world data obtained from Vodafone’s infrastructure are presented, with the superior performance of the newly developed multi-metric predictor being illustrated in comparison to its single-metric counterparts. The dataset utilized in this study, along with the corresponding labeled anomaly dataset, is released under an open data license to facilitate further research in this domain.
