Source-linked AI summary
Resource Allocation in a Network-Based Cloud Computing Environment: Design Challenges
Mohamed Abu Sharkh, Manar Jammal, Abdallah Shami, Abdelkader Ouda
TL;DR
The paper addresses the challenge of designing comprehensive cloud resource-allocation models that jointly account for computation, networking, energy, and other internal and external factors. It organizes these design challenges, identifies research gaps, and provides a reference for developing energy-aware allocation models for cloud data centers. The paper’s supported outcome is a set of discussed challenges and factors rather than a quantitative evaluation.
Problem
Cloud resource-allocation models must represent computational and network resources together while addressing energy consumption and regulatory, geographic, charging, and operational constraints.
Method
The paper introduces and discusses internal and external factors affecting cloud data-center resource-allocation design, including data locality, network reliability, SDN, fault tolerance, portability, consolidation, and VM migration.
Results
The paper identifies design challenges and research gaps as reference points for developing comprehensive energy-aware resource-allocation models for cloud data centers.
Takeaways & Limitations
A complete allocation solution should jointly optimize computational resources, network resources, and energy consumption while accounting for the identified design constraints.
Abstract
from arXiv · showhide
Cloud computing is an increasingly popular computing paradigm, now proving a necessity for utility computing services. Each provider offers a unique service portfolio with a range of resource configurations. Resource provisioning for cloud services in a comprehensive way is crucial to any resource allocation model. Any model should consider both computational resources and network resources to accurately represent and serve practical needs. Another aspect that should be considered while provisioning resources is energy consumption. This aspect is getting more attention from industry and governments parties. Calls of support for the green clouds are gaining momentum. With that in mind, resource allocation algorithms aim to accomplish the task of scheduling virtual machines on data center servers and then scheduling connection requests on the network paths available while complying with the problem constraints. Several external and internal factors that affect the performance of resource allocation models are introduced in this paper. These factors are discussed in detail and research gaps are pointed out. Design challenges are discussed with the aim of providing a reference to be used when designing a comprehensive energy aware resource allocation model for cloud computing data centers.
I. INTRODUCTION
Cloud computing delivers virtualized, scalable resources on demand, but effective service increasingly depends on coordinating computation, networking, and energy use. Prior resource-allocation models typically emphasize processing, network provisioning, or energy efficiency separately, leaving a need for more comprehensive approaches.
- Cloud computing provides virtualized, dynamically scalable computing power, storage, platforms, and services on demand through large data centers.Providers offer these services over the Internet in a pay-as-you-go model.
- Network performance and resource availability can become a cloud bottleneck, so computation alone is insufficient as a competitive advantage.
- Comprehensive resource allocation must manage servers and network resources while satisfying QoS, controlling provider costs, and limiting energy consumption.The system also handles client requests and schedules virtual machines in data centers.
- Previous resource-allocation models: Processing-focused models schedule virtual machines from user requests, sometimes minimizing VM distance, but may assume jobs do not communicate or use only Euclidean network distance.
- Previous resource-allocation models: Network-focused models represent requests as virtual networks and optimize provisioning or revenue, but may omit reservation timing, duration, or added connectivity for existing VMs.
- Previous resource-allocation models: Energy-efficient models commonly consolidate tasks or VMs onto fewer servers and switch unused servers off or into idle states.This approach is modeled as bin packing, with servers treated as bins reaching predefined utilization levels.
II. NETWORK AWARE RA: DESIGN CHALLENGES
A comprehensive cloud resource-allocation model must jointly account for computation, networking, energy, regulatory constraints, and operational challenges. The paper organizes these design challenges into external factors imposed outside allocation and internal factors tied to the allocation algorithm.
- A- A Comprehensive Solution for Network Processing RA: Cloud resource allocation must jointly model computational and network resources because clients need processing capacity and network QoS for application data.The model should address VM allocation, processing-resource modeling, physical distribution, and adequate networking service.
- B- Main Design Challenges: External challenges include geographical and regulatory constraints, client security actions, bandwidth pricing, and distributed-cloud considerations.Data location, encryption or secure wiping, and charging models can constrain placement, performance, QoS, and cost.
- External Challenges 1- Regulative and Geographical Challenges: HIPAA-related location requirements can limit VM placement and data movement while requiring knowledge of data location, copies, modifications, and deletion.The passage states that patient data must not leave US soil.
- 3- SDN Design Challenges inside the DCs: SDN can centralize resource-allocation control, but controller scalability, limited traffic visibility, and controller placement remain internal design challenges.A large data center may generate 20 million flows per second, while current controllers support about 10^5 flows per second in the optimal case.
- B- Main Design Challenges: Cloud resource allocation must also address fault tolerance and portability because hardware, VM, and application failures threaten service continuity, while vendor lock-in complicates VM deployment changes.Clients require applications to remain portable and movable across providers.
III. ENERGY EFFICIENT NETWORK BASED RA
Energy-aware resource allocation must account for rising data-center power use across servers and networks while balancing consolidation, performance, and migration costs.
- Energy motivation: Electricity used by servers doubled from 12 to 23 billion kilowatt hours between 2000 and 2005, while power-related costs approach 50% of data-center operating costs.Growing energy consumption threatens provider expansion and increases environmental concerns.
- Comprehensive energy accounting: A comprehensive energy model should include CPU, memory, disks, power supplies, and network components.Network devices can consume significant aggregate power despite smaller per-device consumption than server racks.
- Network efficiency: Shorter data paths, flow optimization, VM placement, and data-aware scheduling can reduce network energy use.These choices determine whether data stays within a rack, crosses a data center, or travels between geographically distant centers.
- Consolidation trade-offs: Consolidating applications onto fewer servers can save power but may create I/O and network bottlenecks that increase latency and energy use.Concentrated VMs compete for server resources and physical-node links, increasing blocking for data-intensive applications.
- Server state management: Server hibernation and waking introduce latency and power costs, so consolidation should account for those effects and client priorities.Idle servers can consume over 50% of peak power even at very low load.
- Migration trade-offs: VM migration requires weighing server power savings against migration time, network power, and task-completion latency.The trade-off depends on VM size, configuration, and the cost of provisioning new network resources after migration.
IV. CONCLUSION
The paper organizes the design challenges of comprehensive cloud data-center resource allocation across computational, network, and energy concerns. It highlights external constraints and internal challenges that must be addressed in energy-aware models.
- External challenges: The paper identifies regulation, geography, and charging models as external factors affecting data-center resource-allocation design.These factors constrain how resources and data can be allocated across cloud infrastructures.
- Internal challenges: Internal challenges include data locality, reliable data-center networks, SDN, fault tolerance, and portability.The conclusion frames these as design factors for comprehensive resource-allocation models.
- Energy-aware design: Energy-aware allocation faces performance challenges associated with consolidation, VM migration, and server idle-state configuration.The paper presents these challenges as design considerations for future comprehensive models.