Source-linked AI summary
Next Generation Cloud Computing: New Trends and Research Directions
Blesson Varghese, Rajkumar Buyya
TL;DR
Cloud computing is evolving beyond single-provider data centers, creating a need to understand new infrastructure models and architectures. The paper surveys multi-provider and decentralized infrastructure, identifies resulting architectural directions and application impacts, and maps challenges including programming, security, reliability, and sustainability. It concludes that these developments will affect connectivity, data-intensive computing, new services, and self-learning systems.
Problem
Cloud infrastructure is shifting beyond single-provider data centers, but the emerging models, architectures, impacts, and research challenges need to be systematically charted.
Method
The paper discusses changing infrastructure, emerging computing architectures, affected application areas, and challenges for next-generation cloud systems.
Results
The paper identifies impacts on connectivity between people and devices, data-intensive computing, new services, and self-learning systems.
Takeaways & Limitations
Realizing next-generation cloud systems requires addressing challenges in multi-cloud management, edge programming, security, reliability, and sustainable infrastructure.
Takeaways & Limitations
Multi-cloud adoption remains bespoke because provider APIs, abstractions, resources, pricing, and management environments are not unified.
Abstract
from arXiv · showhide
The landscape of cloud computing has significantly changed over the last decade. Not only have more providers and service offerings crowded the space, but also cloud infrastructure that was traditionally limited to single provider data centers is now evolving. In this paper, we firstly discuss the changing cloud infrastructure and consider the use of infrastructure from multiple providers and the benefit of decentralising computing away from data centers. These trends have resulted in the need for a variety of new computing architectures that will be offered by future cloud infrastructure. These architectures are anticipated to impact areas, such as connecting people and devices, data-intensive computing, the service space and self-learning systems. Finally, we lay out a roadmap of challenges that will need to be addressed for realising the potential of next generation cloud systems.
1. Introduction
Cloud computing is moving beyond single-provider data centers toward heterogeneous, multi-provider infrastructure and new computing architectures. The paper charts these trends and research directions for next-generation cloud systems.
- Cloud resources are increasingly drawn from heterogeneous infrastructure across multiple providers rather than a single provider or data center.
- These infrastructure changes are producing new computing architectures that affect societal and scientific areas.
- The paper examines what future cloud computing may look like by charting trends and directions for developing next-generation systems.
- The paper discusses evolving infrastructure, emerging architectures, impact areas, and challenges for developing next-generation cloud systems.
2. Changing Infrastructure
Cloud infrastructure is expanding from centralized, single-provider data centers to multi-cloud, federated, edge, ad hoc, and heterogeneous models. These models offer benefits such as portability, lower latency, and broader resource use, but introduce substantial networking, management, programming, and security challenges.
- Single-provider data-center infrastructure consumes substantial energy, creates centralized failure risks, and can be geographically distant from users.
- Multi-cloud: Multi-cloud applications use resources from multiple providers, but differing APIs, abstractions, hypervisors, pricing, and management tasks require significant programming effort.
- Hybrid Cloud: Hybrid clouds support bursty or known resource demands and sensitive data, but bandwidth, latency, topology, and private-resource management constrain their effectiveness.
- Federated Cloud: Federated clouds can improve interoperability, portability, and service cataloguing, while migration remains difficult because providers differ in abstractions, resources, networks, images, and costs.
- Micro cloud and Cloudlet: Micro clouds decentralize low-power computing toward the network edge, but multi-site networking and integration with existing infrastructure remain unresolved.
- Micro cloud and Cloudlet: Cloudlets extend infrastructure toward mobile users to improve application latency and quality of service while reducing traffic toward distant data centers.
- Ad hoc Cloud: Ad hoc clouds harness underused resources, including smartphones, but require robust management against malicious activity and may drain participating devices’ batteries.
3. Emerging Computing Architectures
The paper presents emerging cloud architectures that extend computing beyond conventional data centers and require changes across the cloud stack. These models include volunteer, fog and mobile edge, serverless, and software-defined computing, each bringing distinct benefits and challenges.
- Overview: Cloud infrastructure is evolving toward new computing models for large-scale applications, requiring changes across cloud-stack abstraction layers.The paper focuses on volunteer computing, fog and mobile edge computing, serverless computing, and software-defined computing.
- Volunteer Computing: Volunteer clouds use spare heterogeneous resources from users or devices, but require incentives, virtualization overhead reduction, security, privacy, and platform integration.Examples include social cloud computing, Cloud@Home payments, and gamification-based incentives.
- Fog and Mobile Edge Computing: Fog computing places resource-constrained computation on edge nodes along the path between users and cloud data centers, reducing latency and improving QoS and QoE.Fog computing complements rather than replaces centralized clouds and may support Internet-of-Things applications.
- Fog and Mobile Edge Computing: Fog architectures can vertically scale workloads across tiers and process data near its source, while Mobile Edge Computing focuses on shared mobile cellular networks.Fog can offload workloads between user devices, edge nodes, and cloud data centers; MEC aims in part to reduce network congestion.
- Fog and Mobile Edge Computing: Fog and mobile edge computing face multi-party management, service-level agreement, security, and privacy challenges.Different parties may own edge nodes, requiring unified management and clearer responsibilities.
- Serverless Computing: Serverless computing executes application functions when events occur rather than keeping an application continuously running on a rented virtual machine.Developers need not manage VM deployment, resource provisioning, scalability, or fault tolerance in the conventional way.
- Serverless Computing: Serverless adoption is expected to grow for edge-connected devices, but programming models must address modularity, control, flexibility, and orchestration trade-offs.The paper identifies programming abstractions and the interaction between traditional external services and serverless services as open challenges.
- Software-Defined Computing: Software-defined computing extends software-defined control from networking to compute and storage across data-center and beyond-data-center resources.The approach is intended to support reconfiguration and adaptation of physical resources while reducing infrastructure configuration complexity.
4. Avenues of Impact
Next-generation cloud architectures are expected to affect connected people and devices, distributed big-data processing, the expanding service space, and self-learning systems. These impacts introduce requirements for security, distributed data management, richer services, and machine-learning capabilities.
- Overview: Distributed, heterogeneous, and ad hoc cloud infrastructure is expected to make cloud systems more ambient, pervasive, and ubiquitous across four impact areas.The paper identifies people and devices, data-intensive computing, the service space, and self-learning systems as key avenues of impact.
- Connecting People and Devices: IoT combines infrastructure, user-device, wearable, and appliance sensors to bring people, devices, and computing closer together.The stated aim is to improve the accuracy and efficiency of actuation and related services.
- Connecting People and Devices: IoT systems require end-to-end security across sensor, wireless, RFID, edge, data-center, and public or private cloud components.Encryption and authentication address outsider attacks, while additional techniques are needed for insider malicious attacks.
- Connecting People and Devices: Future IoT sensing will combine physical, people-centric, and human sensors, creating a need for platforms that handle unstructured data.People-centric sensors localize low-cost environmental sensing to users, while human sensors obtain data directly from people.
- Big Data Computing: Emerging computing models generate Big Data, while centralized storage can create dark data and make later analytics expensive.Processing data before cloud storage is presented as an opportunity to avoid unnecessary movement and storage costs.
- Big Data Computing: Decentralized clouds enable edge processing before storage, but require distributed resource management, vertically scalable analytics, and end-to-end software stacks.These stacks must integrate ad hoc, heterogeneous nodes, micro clouds, cloudlets, and traffic-routing edge nodes.
- Big Data Computing: Unstructured-data analytics must produce accurate and actionable knowledge, with data-to-network-to-knowledge proposed as one approach.The approach transforms unstructured data into structured networks and then into knowledge.
- Service Space: The service space is expanding beyond IaaS, PaaS, and SaaS to include offerings such as Acceleration-as-a-Service, Container-as-a-Service, and Function-as-a-Service.These services target hardware acceleration, container deployment and management, and event-triggered function execution.
5. Research Directions
Next-generation cloud research must address security, application expressivity, marketplaces, reliability, and sustainability across increasingly distributed architectures.
- 5.1. Guaranteeing Enhanced Security: Distributed clouds expand security risks across vertical and horizontal hierarchies, including traffic routed through edge nodes and sensor-integrated privacy concerns.DDoS vulnerabilities also increase as more nodes, devices, virtualization, autoscaling, and multi-tenancy are introduced.
- 5.2. Achieving Expressivity of Applications for Future Clouds: Programming platforms and languages must manage heterogeneous edge resources across multiple hierarchical levels and vendor-specific environments.Emerging models also require limited-startup-latency function execution and alternatives to container-per-function deployment.
- 5.2. Achieving Expressivity of Applications for Future Clouds: Unified environments are needed to address common requirements across emerging architectures while remaining extensible for individual needs.The paper recommends self-managing applications as one way to realize interoperable, application-independent environments.
- 5.3. Developing a Marketplace for Emerging Distributed Architectures: Distributed architectures require marketplaces that account for multiple providers, owners, resources, communication levels, priorities, and transparent agreements among two customer groups.Pricing must articulate who pays across user-device, edge-node, and cloud-server communication paths.
- 5.5. Ensuring Reliability of Cloud Systems: Reliability research must extend replication and disaster recovery to distributed cloud applications while addressing programming effort and recovery costs.Multi-cloud and multi-region architectures are recommended to avoid single points of failure.
- 5.6. Building Sustainable Infrastructure for the Future: Sustainable infrastructure requires geographically distributed coordination, energy- and carbon-aware provisioning, and analysis of trade-offs between energy efficiency, performance, and SLA compliance.Energy efficiency has been proposed as a QoS metric, but stricter VM policies may risk SLA violations.
6. Summary
Cloud computing is moving from single-provider, centralized data centers toward multi-provider and decentralized infrastructure with new computing models. The paper surveys resulting architectures, anticipated application areas, and research challenges for realizing next-generation cloud systems.
- Changing infrastructure: Cloud infrastructure is trending toward multiple providers and computing distributed away from centralized data-center resources.This contrasts with traditional single-provider cloud offerings and motivates new computing models.
- Emerging computing models: The paper considers ad hoc clouds based on voluntarily provided resources, edge computing for mobile and online applications, idle-aware cloud payment, and resilient software-defined distributed computing.These are presented as emerging computing models highlighted by the paper.
- Impacted areas: Next-generation clouds are anticipated to improve connectivity, support data-intensive computing, expand services such as containers and functions, and contribute to self-learning systems.These impacts span connections between people and devices, data processing, service delivery, and convergent research areas.
- Challenges: Realizing these systems requires research directions in security, application expressivity, efficient management, and sustainable system development.The paper frames these directions as future challenges rather than completed solutions.