Performance Modelling and Simulation of Three-Tier Applications in Cloud and Multi-Cloud Environments

Abstract

A significant number of Cloud applications follow the 3-tier architectural pattern. Many of them serve customers worldwide and must meet non-functional requirements such as reliability, responsiveness and Quality of Experience (QoE). Thus the flexibility and scalability offered by clouds make them a suitable deployment environment. Latest developments show that using multiple clouds can further increase an application's reliability and user experience to a level that has not been achievable before. However, the research in scheduling and provisioning 3-tier applications in clouds and across clouds is still in its infancy.To foster the research efforts in the area, we propose an analytical performance model of 3-tier applications in Cloud and Multi-Cloud environments. It takes into account the performance of the persistent storage and the heterogeneity of cloud data centres in terms of Virtual Machine (VM) performance. Furthermore, it allows for modelling of heterogeneous workloads directed to different data centres. Based on our model, we have extended the CloudSim simulator, through which we validate the plausibility of our approach. The conducted experiments with the RUBiS workload show that the predicted performance characteristics by the simulation approximate well those of the modelled system.

Workload modeling for resource usage analysis and simulation in cloud computing

“…However, different types of users directly impact the cloud workload as observed by the cloud provider. In view of this, Moreno et al [3] and Grozev and Buyya [9] created models based on resource utilization and users' behavioral patterns. Moreno et al [3] proposed a new approach for characterization of the Google trace log in the context of both user and task in order to derive a model to capture resource utilization patterns.…”

“…Grozev and Buyya [9] also made use of the Rice University Bidding System (RUBiS) workload and implemented the model in the CloudSim. The Central Processing Unit (CPU) load is modeled in terms of average number of instructions required by a user session.…”

“…However, recent works have been focusing on provisioning and scheduling techniques for batch applications, while cloud resource management in interactive applications has not received much attention so far [9]. In view of this, we utilize the widely used RUBiS [10] benchmark in order to evaluate the impact of the user in the resource consumption patterns.…”

CEPSim: A Simulator for Cloud-Based Complex Event Processing

“…Because of its flexibility, CloudSim has originated many different extensions in the literature [6], [21], [22]. Garg and Buyya [6] created NetworkCloudSim, which extends CloudSim with a three-tier network model.…”

“…Guérout et al [21], on the other hand, focussed on implementing the DVFS model on CloudSim. Finally, Grozev and Buyya [22] presented a model for three-tier Web applications and incorporated it into CloudSim. These extensions are orthogonal to those presented in this paper because they do not focus on CEP.…”