Adaptable TeaStore

2024-12-20

arXiv.org

AI Summary

Existing cloud-native benchmarks lack support for dynamic adaptability—such as service scaling, sudden load shifts, and dependency failures—hindering systematic evaluation of adaptive architectures. This paper introduces TeaStore, a configurable microservice benchmarking platform. It is the first to incorporate a service criticality classification mechanism (mandatory vs. optional), enabling runtime architectural pruning and plug-and-play integration of third-party services. Through containerized deployment, declarative configuration, and standardized API gateway integration, TeaStore ensures fault isolation and graceful degradation. This design significantly enhances system stability under dynamic perturbations and improves experimental reproducibility. As a result, TeaStore establishes a more production-realistic baseline for evaluating cloud resource scheduling, elasticity modeling, and resilience assessment—bridging the gap between benchmarking environments and real-world operational conditions.

Technology Category

• Machine Learning: Scalability of ML Systems
• Data Mining & Knowledge Management: Scalability, Parallel & Distributed Systems
• Search and Optimization: Algorithm Configuration

Application Category

• Systems and Infrastructure for Web, Mobile and WoT: Web performance, measurement, and characterization
• Security and Privacy: Large-scale security measurements
• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics

Abstract

Adaptability is a fundamental requirement for modern Cloud software architectures to ensure robust performance in the face of diverse known and unforeseen events inherent to distributed systems. State-of-the-art Cloud systems frequently adopt microservices or serverless architectures. Among these, TeaStore is a recognised microservice reference architecture that offers a benchmarking framework for modelling and resource management techniques. However, TeaStore's original configuration lacks the flexibility necessary to address the varied scenarios encountered in real-world applications. To overcome this limitation, we propose an enhanced variant of TeaStore that distinguishes between mandatory and optional services while incorporating third-party service integration. Core services such as WebUI, Image Provider, and Persistence are designated as mandatory to maintain essential functionality, whereas optional services, such as Recommender and Auth, extend the architecture's feature set. We outline the design and configuration possibilities of this adaptable TeaStore variant, aimed at enabling a broader spectrum of configurability and operational resilience.

Problem

• Studies adaptation mechanisms for cloud-native systems
• Addresses service outages and traffic surges dynamically
• Explores reconfiguration policies in diverse operational scenarios

Innovation

• Introduces Adaptable TeaStore with mandatory and optional services
• Supports multiple component versions and external outsourcing
• Provides local cache mechanisms for performance and resilience