Transformation-Based Monetary Cost
Optimizations for Workflows in the Cloud
Abstract
Recently, performance and monetary cost optimizations for workflows from various applications in the cloud have become ahot research topic. However, we find that most existing studies adopt ad hoc optimization strategies, which fail to capture the keyoptimization opportunities for different workloads and cloud offerings (e.g., virtual machines with different prices). This paper proposesToF, a general transformation-based optimizationframework for workflows in the cloud. Specifically, ToF formulates six basic workflowtransformation operations. An arbitrary performance and cost optimization process can be represented as a transformation plan (i.e., asequence of basic transformation operations). All transformations form a huge optimization space. We further develop a cost modelguided planner to efficiently find the optimized transformation for a predefined goal (e.g., minimizing the monetary cost with a givenperformance requirement). We develop ToF on real cloud environments including Amazon EC2 and Rackspace. Our experimentalresults demonstrate the effectiveness of ToF in optimizing the performance and cost in comparison with other existing approaches.
Existing Scheme
We have witnessed many scientific applications partially or entirely shifting from traditional computing platforms (e.g., grid) to the cloud. Due to the pay-as-you-go computational behavior, performance and (monetary) cost optimizations have recently become a hot research topic for workflows in the cloud. A lot of scheduling and optimization approaches have been developed.Despite of a lot of research efforts in this area, performance and cost optimizations of workflows in the cloud are still a non-trivial task, because of the following complicated and inter-connected factors. First, users have different requirements on performance and cost. Some existing studies have focused on minimizing the cost while satisfying the performance requirement, some are aiming to optimize performance for a given budget and others are considering the trade-off between performance and monetary cost. Second, different cloud offerings result in significantly different cost structures for running the workflow. Even from the same cloud provider, there are multiple types of virtual machines (or instances) with different prices and computing capabilities. Third, workflows have very complicated structures and different computation/IO characteristics, as observed in the studies. All those factors call for a general and effective approach for performance and cost optimizations.
To address the limitations of current approaches, we propose ToF, a transformation-based optimization framework for optimizing the performance and cost of workflows in the cloud. ToF models the cost and performance optimizations of worklows as transformations. We categorize the transformation operations into two kinds, namely main schemes and auxiliary schemes. The main schemes reduce monetary cost while auxiliary schemes transform workflows into a DAG that is suitable for main schemes to perform cost reduction. We further develop a cost model guided planner to help users to efficiently and effectively choose the cost-effective transformation. Moreover, we develop heuristics (e.g., iteratively choosing the cost-effective main scheme and auxiliary scheme) to reduce the runtime overhead of the optimization process.
Proposed scheme
Three design principles in mind, we proposeToF, a transformation-based optimization framework foroptimizing the performance and cost of workflows in thecloud. A workflow is generally modeled as a directed acyclicgraph (DAG) of tasks. ToF guides the scheduling ofeach task in the workflow, including which instance toassign to and when to start execution.The searching space for an optimal transformationsequence is huge. Second, the optimization is an online processand should be lightweight. We should find a good balancebetween the quality of the transformation sequence andthe runtime overhead of the planner. Performance and monetary cost optimizations for runningworkflows from different applications in the cloud havebecome a hot and important research topic. Those issues include relatively limited cross-cloud networkbandwidth and lacking of cloud standards amongcloud provider.
CONCLUSION
Performance and monetary cost optimizations for runningworkflows from different applications in the cloud havebecome a hot and important research topic. However, mostexisting studies fail to offer general optimizations to captureoptimization opportunities in different user requirements,cloud offerings and workflows. To bridge this gap, we proposea workflow transformation-based optimization frameworknamely ToF. We formulate the performance and costoptimizations of workflows in the cloud as transformation
and optimization. The two components are designed to beextensible for user requirements on performance and cost,cloud offerings and workflows. Particularly, we formulatesix basic transformation operations. We further develop acost model guided planner to efficiently and effectively findthe suitable transformation sequence for the given performanceand cost goal. We evaluate our framework usingreal-world scientific workflow applications and comparewith other state-of-the-art scheduling algorithms. Resultsshow our framework outperforms the state-of-the-art Autoscalingalgorithm by 30 percent for monetary cost optimization,and by 21 percent for the execution time optimization.Moreover, the planner is lightweight for online optimization
in the cloud environments. As for future work, we considerToF on multiple clouds. Still, there are many practical andchallenging issues for current multi-cloud environments. Those issues include relatively limited cross-cloud network
bandwidth and lacking of cloud standards amongcloud providers.
System Configuration:-
Hardware Configuration:-
Processor-Pentium –IV
Speed- 1.1 Ghz
RAM- 256 MB(min)
Hard Disk- 20 GB
Key Board- Standard Windows Keyboard
Mouse- Two or Three Button Mouse
Monitor- SVGA
Software Configuration:-
Operating System: Windows XP
Programming Language: JAVA
Java Version: JDK 1.6 & above.
Further Details Contact: A Vinay 9030333433, 08772261612
Email: |