Name
Abstract | ||
|---|---|---|
In a big data environment, elastic clusters are introduced both to fulfill Service Quality Requirements (SQRs) and to save system energy consumption. In most elastic clusters, one or some features of system status are used as the hint for resource scaling. For example, request arrival rate is a widely used feature. Besides, system utilization, mean service time (mean request size) and cache queue length are also common features. However, in elastic clusters where the SQR is referred to the response time of requests, these features are not good choices. For the reason that these features can not well reflect the status of System Service Quality (SSQ). Consequently, it's difficult to achieve an accurate resource scaling, and the system energy-saving efficiency and SSQ can not be well maintained. In this paper, we show that in such type of elastic clusters, the outstanding hint for resource scaling is the request response time. So we propose a resource scaling method which scales resources leveraging the hint of response time. More specifically, our method uses a circular queue to maintain the response time of recent requests, then the mean value is calculated as the hint of response time. Because this hint directly and exactly reflects the status of SSQ, our method can significantly promote system performance on both the SSQ and the energy-saving efficiency. The experimental results validate our point of view and show that our method achieves a good performance by using this hint. |
Year | DOI | Venue |
|---|---|---|
2019 | 10.1109/iThings/GreenCom/CPSCom/SmartData.2019.00076 | 2019 International Conference on Internet of Things (iThings) and IEEE Green Computing and Communications (GreenCom) and IEEE Cyber, Physical and Social Computing (CPSCom) and IEEE Smart Data (SmartData) |
Keywords | Field | DocType |
elastic clusters,resource scaling,system service quality,energy saving | Service quality,Computer science,Cache,Server,Queue,Circular buffer,Algorithm,Response time,Scaling,Request–response | Conference |
ISBN | Citations | PageRank |
978-1-7281-2981-5 | 0 | 0.34 |
References | Authors | |
16 | 1 | |