[Paper Review] PhoenixCloud: Provisioning Resources for Heterogeneous Workloads in Cloud Computing

Jianfeng Zhan, Lei Wang+3 more|arXiv (Cornell University)|Jun 8, 2010

Cloud Computing and Resource Management34 references3 citations

TL;DR

PhoenixCloud proposes a coordinated resource provisioning system for heterogeneous workloads—specifically parallel batch jobs and Web services—in both private and public Cloud environments. By dynamically managing resource allocation through adaptive policies, it reduces cluster configuration size by 40% in private Clouds and cuts peak resource consumption by up to 31% compared to EC2+RightScale in public Clouds.

ABSTRACT

As more and more service providers choose Cloud platforms, which is provided by third party resource providers, resource providers needs to provision resources for heterogeneous workloads in different Cloud scenarios. Taking into account the dramatic differences of heterogeneous workloads, can we coordinately provision resources for heterogeneous workloads in Cloud computing? In this paper we focus on this important issue, which is investigated by few previous work. Our contributions are threefold: (1) we respectively propose a coordinated resource provisioning solution for heterogeneous workloads in two typical Cloud scenarios: first, a large organization operates a private Cloud for two heterogeneous workloads; second, a large organization or two service providers running heterogeneous workloads revert to a public Cloud; (2) we build an agile system PhoenixCloud that enables a resource provider to create coordinated runtime environments on demand for heterogeneous workloads when they are consolidated on a Cloud site; and (3) A comprehensive evaluation has been performed in experiments. For two typical heterogeneous workload traces: parallel batch jobs and Web services, our experiments show that: a) in a private Cloud scenario, when the throughput is almost same like that of a dedicated cluster system, our solution decreases the configuration size of a cluster by about 40%; b) in a public Cloud scenario, our solution decreases not only the total resource consumption, but also the peak resource consumption maximally to 31% with respect to that of EC2 +RightScale solution.

Motivation & Objective

• Address the challenge of efficiently provisioning resources for heterogeneous workloads—such as batch jobs and Web services—across different Cloud deployment models.
• Overcome limitations of traditional runtime environments that support only homogeneous workloads, like OpenPBS or Oceano.
• Design a system that enables on-demand creation of coordinated runtime environments to improve resource utilization and service quality.
• Provide a unified solution for both private Clouds (dedicated infrastructure) and public Clouds (outsourced infrastructure) with minimal performance degradation.
• Achieve high resource efficiency while maintaining acceptable job turnaround times through adaptive resource allocation policies.

Proposed method

• Propose a coordinated resource provisioning framework that dynamically allocates resources based on workload characteristics and demand patterns.
• Introduce a policy-based system with parameters B (baseline), U (utilization), V (vulnerability), and G (elasticity of release) to tune resource allocation behavior.
• Implement the PhoenixCloud system to create and manage coordinated runtime environments on demand, enabling shared resource pools across heterogeneous workloads.
• Use workload traces from NASA iPSC, SDSC BLUE, and WorldCup to simulate real-world scenarios and evaluate performance under varying configurations.
• Apply two key policies: FB (Fixed Baseline) and FLB-NUB (Fixed and Low Baseline with Non-Uniform Bounds), to balance resource consumption and job turnaround time.
• Integrate with existing Cloud infrastructure (e.g., EC2) and compare performance against EC2+RightScale to validate efficiency gains.

Experimental results

Research questions

• RQ1 Can coordinated resource provisioning significantly reduce resource consumption and configuration size for heterogeneous workloads in private Clouds?
• RQ2 How does PhoenixCloud’s dynamic resource allocation compare to EC2+RightScale in terms of peak and total resource usage in public Clouds?
• RQ3 What are the trade-offs between resource efficiency and job turnaround time when consolidating batch jobs and Web services in shared Cloud environments?
• RQ4 How do configurable parameters (B, U, V, G) affect system performance and resource utilization in heterogeneous workload consolidation?
• RQ5 To what extent can coordinated runtime environments improve resource utilization without degrading service quality for diverse workloads?

Key findings

• In private Cloud scenarios, PhoenixCloud reduces cluster configuration size by approximately 40% compared to a dedicated cluster system while maintaining similar throughput.
• For public Cloud deployments, PhoenixCloud reduces peak resource consumption by up to 31% and total resource consumption compared to the EC2+RightScale baseline.
• The FLB-NUB policy achieves a favorable trade-off, reducing both total and peak resource usage while increasing average turnaround time by only a small margin compared to EC2+RightScale.
• The parameter G (elasticity of resource release) has a significant impact on average turnaround time; a value of 0.5 provides a good balance between responsiveness and resource savings.
• Parameters U and V have minimal impact on key metrics when set within the range of 1.0–2.0 and 0.1–0.5, respectively, suggesting robustness to moderate configuration changes.
• Management overhead increases with shorter leasing time units, leading to the selection of 60 minutes as the optimal time unit for leasing resources in PhoenixCloud and EC2.