Source-linked AI summary
InterCloud: Utility-Oriented Federation of Cloud Computing Environments for Scaling of Application Services
Rajkumar Buyya, Rajiv Ranjan, Rodrigo N. Calheiros
TL;DR
Federated Clouds address the difficulty of scaling composite applications across geographically distributed providers while meeting QoS and efficiency goals. The paper proposes an architectural federation of coordinators, brokers, and an exchange, with resource allocation and migration supporting cross-domain service integration. Its conclusion presents this framework as a way to manage components in large, dynamic federated environments while protecting customers with guaranteed QoS.
Problem
Composite applications require flexible service-to-resource mapping across distributed Clouds, while integration, interoperability, and VM migration remain constrained by uncertain resources and existing bottlenecks.
Method
The paper proposes a service-oriented federation architecture combining Cloud Coordinators, Brokers, and an Exchange for cross-domain scheduling, allocation, and workload migration.
Results
The framework is presented as facilitating federated management of system components and protecting customers with guaranteed quality of service in large, dynamic environments.
Takeaways & Limitations
Federated Cloud management aims to support elastic application services and utility-oriented access across distributed infrastructure.
Takeaways & Limitations
VM migration across federated service domains faces a serious bottleneck because the environments do not yet provide shared networking and storage.
Abstract
from arXiv · showhide
Cloud computing providers have setup several data centers at different geographical locations over the Internet in order to optimally serve needs of their customers around the world. However, existing systems do not support mechanisms and policies for dynamically coordinating load distribution among different Cloud-based data centers in order to determine optimal location for hosting application services to achieve reasonable QoS levels. Further, the Cloud computing providers are unable to predict geographic distribution of users consuming their services, hence the load coordination must happen automatically, and distribution of services must change in response to changes in the load. To counter this problem, we advocate creation of federated Cloud computing environment (InterCloud) that facilitates just-in-time, opportunistic, and scalable provisioning of application services, consistently achieving QoS targets under variable workload, resource and network conditions. The overall goal is to create a computing environment that supports dynamic expansion or contraction of capabilities (VMs, services, storage, and database) for handling sudden variations in service demands. This paper presents vision, challenges, and architectural elements of InterCloud for utility-oriented federation of Cloud computing environments. The proposed InterCloud environment supports scaling of applications across multiple vendor clouds. We have validated our approach by conducting a set of rigorous performance evaluation study using the CloudSim toolkit. The results demonstrate that federated Cloud computing model has immense potential as it offers significant performance gains as regards to response time and cost saving under dynamic workload scenarios.
2 Manjrasoft Pty Ltd, Australia
The passage identifies the School of Computer Science and Engineering at the University of New South Wales in Sydney, Australia.
- The affiliation is the School of Computer Science and Engineering.
- The institution is the University of New South Wales.
- The location is Sydney, Australia.
1. Introduction
The paper motivates federated Clouds for elastic, QoS-aware application services across distributed providers and identifies architectural and technical requirements for realizing that vision.
- Cloud computing provides on-demand access to virtualized application services without regard to where they are hosted.
- Social Networking Web Applications: Composite social-networking applications may span hundreds of services and dozens of geographically distributed Cloud data centers, requiring dynamic scaling as workload locality changes.
- Existing systems lack seamless automatic scaling across geographically distributed data centers, making it difficult to select hosting locations and meet consumers’ QoS expectations.
- Federated providers should lease resources across Clouds, negotiate SLA-driven market prices, and deliver reliable, cost-effective, QoS-aware services.
- Research Issues: Key research challenges include predicting service behavior, mapping services to resources, optimizing economic deployment decisions, integrating on-premises systems, and monitoring distributed components scalably.
- Overall Vision: The proposed federation combines Cloud Coordinators, Brokers, and an Exchange with market-oriented provisioning and energy-aware placement and migration policies.
2. State-of-the-art in Cloud Provisioning
Existing public Cloud platforms provide varied provisioning, monitoring, load-balancing, and auto-scaling capabilities, but providers still lack globally distributed coverage needed to meet QoS expectations for all users.
- Public Cloud platforms expose differing capabilities for monitoring, managing, and provisioning resources and application services.The paper compares Amazon Web Services, Microsoft Azure, Google AppEngine, Manjrasoft Aneka, Eucalyptus, and GoGrid.
- AWS combines Elastic Load Balancing, Auto Scaling, and CloudWatch to distribute workload and adjust EC2 capacity as demand changes.
- Aneka supports distributed application deployment, resource provisioning, monitoring, pricing, accounting, and QoS/SLA services across private, public, and hybrid Clouds.
- Eucalyptus uses Cluster Controllers to monitor instance state, select available servers, and provision incoming requests, but still lacks some capabilities compared with AWS.
- Google App Engine automatically manages load balancing, service provisioning, and auto-scaling, but supports only web-hosting applications.
- Because no single provider covers all global locations, SaaS providers may struggle to meet QoS expectations and therefore prefer federated infrastructures combining multiple Clouds.
3. System Architecture and Elements of InterCloud
InterCloud federates geographically and administratively distributed Cloud resources through brokers, coordinators, and a Cloud Exchange, enabling market-based, QoS-aware provisioning and workload placement.
- System Architecture: The architecture integrates brokering and coordination services so distributed Cloud storage and compute resources appear as a single resource-leasing abstraction.
- Cloud Exchange: The Cloud Exchange matches brokered infrastructure demand with supply published by Cloud Coordinators and supports Cloud-service trading.
- Cloud Broker: Each client instantiates a Cloud Broker that establishes service contracts with Cloud Coordinators through Cloud Exchange trading functions.
- Cloud Coordinator: The Cloud Coordinator manages enterprise-Cloud membership in the federation through market-based trading and negotiation protocols.
- Scheduling and Allocation: Scheduling and allocation assign virtual machines according to user QoS targets and Cloud energy-management goals, using application, sensor, and policy feedback.
- Market and Policy Engine: The coordinator’s market and policy functions support SLAs, pricing, accounting, and billing based on resource supply, demand, and actual usage.
- Resource Management Challenges: Federated provisioning must address conflicting QoS targets, workload variation, monitoring, discovery, application composition, and VM migration across Cloud domains.The paper identifies multi-dimensional optimization, dynamic monitoring, and cross-domain virtualization as central design challenges.
- Sensing and Monitoring: Sensors monitor node power consumption, heat dissipation, and utilization to provide dynamic information for Cloud coordination.
4. Early Experiments and Preliminary Results
The experiments use CloudSim to compare federated and non-federated cloud configurations, including load migration and dynamic public-cloud resource leasing. Federation reduces application execution time, while additional resources reduce makespan but increase processing cost.
- Federated performance evaluation: CloudSim models three federated cloud providers, a Cloud Broker, Sensors, and a Cloud Coordinator that uses availability statistics for load-migration decisions.The federation is compared with independent data centers that process all submitted workload locally.
- Federated performance evaluation: The federated configuration lets data centers handle peak demands by migrating excess load to the least-loaded clouds.In the non-federated configuration, each provider executes its workload locally.
- Federated performance evaluation: More than 50% lower average Cloudlet turn-around time and 20% better application makespan are achieved with federated infrastructure.These results are reported for the comparison summarized in Table 2.
- Federated provisioning strategy: Increasing resources at the stated rate reduces job makespan at the same rate, but processing cost increases significantly at higher rates.The experiment evaluates a private cloud supplemented by dynamically leased public-cloud resources.
- Federated provisioning strategy: Leasing public-cloud resources is reported as cheaper than buying and maintaining extra capacity that would remain idle most of the time.The authors characterize the extra public-cloud cost as acceptable because peak demand occurs only occasionally.
5. Conclusions and Future Directions
The paper concludes that federating distributed clouds can support elastic application services with improved scalability, flexibility, and management. It identifies adaptive, model-driven provisioning and energy-aware resource management as important directions for future work.
- Conclusions: Integrating distributed clouds through federation is presented as critical for composing and deploying elastic application services.The proposed framework targets federated management of services and resources in dynamic environments.
- Conclusions: The framework aims to improve scalability, flexibility, and simplicity in managing and delivering services across federated clouds.Its components address both service management and resource management.
- Future directions: Future work will develop model-driven provisioning that relates high-level performance targets to low-level controllable parameters and observables.Planned techniques include multi-class queuing models, linear regression time series, and real-time monitoring.
- Future directions: Energy-efficient management remains important because growing applications and data require increasingly larger servers and disks.The paper frames reduced energy usage as necessary for sustainable future cloud computing.
- Future directions: Market-based utility models are proposed as a way to combine revenue optimization with energy-efficient utilization of cloud infrastructure.The approach accepts requests that can be fulfilled among competing demands.