Abstract | ||
---|---|---|
This work addresses performance testing for monitoring mass quantities of large-dataset measurements in infrastructure-as-a-Service (IaaS). Physical resources are not virtualized in sharing dynamic clouds; thus, shared resources compete for access to system resources. This competition introduces significant new challenges when assessing the performance of IaaS. A bottleneck may occur if one system resource is critical to IaaS; this may shut down the system and services, which would reduce the workflow performance by a large margin. To protect against bottlenecks, we propose CloudPT, a performance test management framework for IaaS. CloudPT has many advantages: (I) high-efficiency detection; (II) a unified end-to-end feedback loop to collaborate with cloud-ecosystems management; and (III) a troubleshooting performance test. This paper shows that CloudPT efficiently identifies and detects bottlenecks with a minimal false-positive rate (u003c13%) and it correlates high accuracy using the failure of a host virtual machine (host VM) to start-up with both cloud illustrative batches and transactional workloads such as the Spark, and Kafka framework for a data partitioning and collecting events on an each server. In a framework based on a trace case study, CloudPT diagnosed performance bottlenecks in 20 s with a precision rate of 86%, confirming its real-time efficiency. |
Year | Venue | Field |
---|---|---|
2018 | ICA3PP | Troubleshooting,Bottleneck,Test management,Virtual machine,Spark (mathematics),Computer science,Feedback loop,Workflow,Cloud computing,Distributed computing |
DocType | Citations | PageRank |
Conference | 0 | 0.34 |
References | Authors | |
14 | 3 |
Name | Order | Citations | PageRank |
---|---|---|---|
Ameen Alkasem | 1 | 0 | 0.34 |
Hongwei Liu | 2 | 376 | 63.93 |
De-Cheng Zuo | 3 | 86 | 18.87 |