Gueyoung Jung

Synchronous Parallel Processing of Big-Data Analytics Services to Optimize Performance in Federated Clouds

Parallelization of big-data analytics services over a federation of heterogeneous clouds has been considered to improve performance. However, contrary to common intuition, there is an inherent tradeoff between the level of parallelism and the performance for big-data analytics principally because of a significant delay for big-data to get transferred over the network. The data transfer delay can be comparable or even higher than the time required to compute data. To address the aforementioned tradeoff, this paper determines: (a) how many and which computing nodes in federated clouds should be used for parallel execution of big-data analytics; (b) opportunistic apportioning of big-data to these computing nodes in a way to enable synchronized completion at best-effort performance; and (c) sequence of apportioned, different sizes of big-data chunks to be computed in each node so that transfer of a chunk is overlapped as much as possible with the computation of the previous chunk in the node. In this regard, Maximally Overlapped Bin-packing driven Bursting (MOBB) algorithm is proposed, which improve the performance by up to 60% against existing approaches.

Cloud Capability Estimation and Recommendation in Black-Box Environments Using Benchmark-Based Approximation

As cloud computing has become popular and the number of cloud providers has proliferated over time, the first barrier to cloud users will be how to accurately estimate performance capabilities of many different clouds and then, select a right one for given complex workload based on estimates. Such cloud capability estimation and selection can be a big challenge since most clouds can be considered as black-boxes to cloud users by abstracting underlying infrastructures and technologies. This paper describes a cloud recommender system to recommend an optimal cloud configuration to users based on accurate estimates. To achieve this, our system generates the capability vector that consists of relative performance scores of resource types (e.g., CPU, memory, and disk) estimated for given user workload using benchmarks. Then, a search algorithm has been developed to identify an optimal cloud configuration based on these collected capability vectors. Experiments show our approach accurately estimate the performance capability (less than 10% error) while scalable in large search space.

An economic model for green cloud

A novel economic model for cloud-based services is presented that: (i) transparently presents energy demands (of services) to the customers in a simple abstract form, called green point, which is understandable to any general user; (ii) provides economic incentives (through dynamic discounts) as motivations for customers to select greener configuration; and (iii) offers service prices to customers such that the profit of cloud vendor is maximized while providing the discounts. Price is differentiated for different classes of customers (e.g. gold, silver, and bronze) and dynamic based on posterior distribution on resource demand considering both current demand and willingness toward green configuration. The model enables a paradigm shift in cloud service offering that provides higher transparency and control knobs to users for greener configuration. Preliminary results indicate higher profit using the proposed model compared to static pricing in existing pay-per-use service offerings.

Optimal Time-Cost Tradeoff of Parallel Service Workflow in Federated Heterogeneous Clouds

Federated cloud enables a workflow to be deployed in multiple private and public clouds. By facilitating external cloud-based services to execute sub-tasks of the workflow, service workflow owners can reduce the cost of executing the workflow, while meeting a performance requirement, since those cloud-based services can be more cost efficient and have better performance than internal ones. However, due to inter-dependencies between sub-tasks, the complexity of the workflow, and the heterogeneity of clouds, it is a challenge to achieve the optimal tradeoff between cost and performance. This paper presents a novel workflow scheduler designed to achieve the optimal end-to-end execution time and cost when deploying such complex workflows in heterogeneous computing nodes in clouds. Specifically, our scheduling algorithm addresses the tradeoff between the execution cost, the computing time, and the data transfer delay between sub-tasks. Our scheduler can handle complex workflows that contain recursively paralleled sub-flows caused by branch and merging sub-tasks. Experiments indicate that our scheduler can efficiently compute the near optimal deployment compared with greedy and evolutionary algorithms for both end-to-end execution time and corresponding cost.

Towards Simplifying and Automating Business Process Lifecycle Management in Hybrid Clouds

Business Process Management (BPM) software provides visibility into business processes in organizations of all sizes and helps increase process efficiency continuously. However, the time and effort involved in modeling, deploying and executing a business process is tremendous and as a result organizations struggle to agilely adapt business processes to dynamic business requirements. On the other hand, the growing popularity of cloud computing poses opportunities and challenges on how business processes can leverage resource outsourcing and elasticity. In light of the above, this paper presents a business process management platform that assists business analysts lacking necessary programming expertise by automating manual steps and providing guidance and recommendations to quickly and efficiently design, implement, deploy and execute business processes in a hybrid cloud environment.