Public Cloud Users Research Articles

Online cloud resource prediction via scalable window waveform sampling on classified workloads

Accurate prediction on the utilization of cloud resources is increasingly important for public cloud users, as it relates to the reasonable reservation of resources for minimizing the usage costs. However, the existing relevant approaches fail to predict the usage amount of cloud resources on the basis of the requested workloads of users’ applications, and the characteristics of changing workload data are rarely considered for the real-time prediction. To address these challenges, we propose an online cloud resource prediction model (OCRPM) to timely predict the proper resource usage amount. Firstly, all of the requested workloads are classified into three types of waveform trend patterns using the trend degree (TD). Next, a scalable window waveform sampling method (SWWS) on the classified patterns is devised to extend the suitable workload waveform interval window for supporting the subsequent high accurate prediction on the cloud resources. Finally, an optimal error gradient boosting regression (OEGBR) algorithm is given to train the data model and to predict the reasonable cloud resource usage amount in light of the requested workloads. The simulation results indicate that the proposed method can adjust the suitable workload waveform sampling window, and achieve higher prediction accuracy than the state-of-the-art relevant approaches and existing statistical learning models.

The Limit of Horizontal Scaling in Public Clouds

Public cloud users are educated to practice horizontal scaling at the application level, with the assumption that more processing capacity can be achieved by adding nodes into the server fleet. In reality, however, applications—even those specifically designed to be horizontally scalable—often face unpredictable scalability issues when running at scale. In this article, we study the limit of horizontal scaling in public clouds by identifying sources of such limitations and quantitatively measuring their impact on processing capacity. To this end, we develop ScaleBench as a distributed and parallel cloud-scale testing framework and propose a capacity degradation index (CDI) to describe the level of capacity degradation observed in our benchmark studies. We have conducted extensive experiments in four real public clouds to identify possible bottlenecks in compute, block storage, networking, and object storage. Further, we carry out large-scale experiments with a real-life video transcoding application on worker fleets with up to 3200 vCPU cores. Our experimental results provide the quantitative evidence on the limit of horizontal scaling in public clouds. This helps cloud users make better design decisions on horizontally scalable applications.

Securing Cryptographic Keys in the Cloud: A Survey

Public cloud users are increasingly interested in securing their data and network traffic by using outsourced encryption or decryption services (cryptography as a service). However, many concerns have been raised about trusting the outsourced service's memory with sensitive information like the user's cryptographic keys. This survey reviews techniques for securing data owners' cryptographic keys in a public cloud, preserving them from any inquisitive actions on the part of cotenants (fellow data owners, domain administrators, or other third parties) and the cloud provider itself. The aim is to provide readers with a comparative in-depth analysis of available techniques and point to open research problems.

Characterizing spot price dynamics in public cloud environments

The surge in demand for utilizing public Cloud resources has introduced many trade-offs between price, performance and recently reliability. Amazon’s Spot Instances (SIs) create a competitive bidding option for public Cloud users at lower prices without providing reliability on services. It is generally believed that SIs reduce monetary cost to the Cloud users, however it appears from the literature that their characteristics have not been explored and reported. We believe that characterization of SIs is fundamental in the design of stochastic scheduling algorithms and fault tolerant mechanisms in public Cloud environments for the spot market. In this paper, we have done a comprehensive analysis of SIs based on one year price history in four data centers of Amazon’s EC2. For this purpose, we have analyzed all different types of SIs in terms of spot price and the inter-price time (time between price changes) and determined the time dynamics for spot price in hour-in-day and day-of-week. Moreover, we have proposed a statistical model that fits well these two data series. The results reveal that we are able to model spot price dynamics as well as the inter-price time of each SI by a mixture of Gaussians distribution with three or four components. The proposed model is validated through extensive simulations, which demonstrate that our model exhibits a good degree of accuracy under realistic working conditions.