Rasha Osman

CloudScope: Diagnosing and Managing Performance Interference in Multi-tenant Clouds

Virtual machine consolidation is attractive in cloud computing platforms for several reasons including reduced infrastructure costs, lower energy consumption and ease of management. However, the interference between co-resident workloads caused by virtualization can violate the service level objectives (SLOs) that the cloud platform guarantees. Existing solutions to minimize interference between virtual machines (VMs) are mostly based on comprehensive micro-benchmarks or online training which makes them computationally intensive. In this paper, we present CloudScope, a system for diagnosing interference for multi-tenant cloud systems in a lightweight way. CloudScope employs a discrete-time Markov Chain model for the online prediction of performance interference of co-resident VMs. It uses the results to optimally (re)assign VMs to physical machines and to optimize the hypervisor configuration, e.g. the CPU share it can use, for different workloads. We have implemented CloudScope on top of the Xen hypervisor and conducted experiments using a set of CPU, disk, and network intensive workloads and a real system (MapReduce). Our results show that CloudScope interference prediction achieves an average error of 9%. The interference-aware scheduler improves VM performance by up to 10% compared to the default scheduler. In addition, the hypervisor reconfiguration can improve network throughput by up to 30%.

Understanding, modelling, and improving the performance of web applications in multicore virtualised environments

As the computing industry enters the Cloud era, multicore architectures and virtualisation technologies are replacing traditional IT infrastructures. However, the complex relationship between applications and system resources in multicore virtualised environments is not well understood. Workloads such as web services and on-line financial applications have the requirement of high performance but benchmark analysis suggests that these applications do not optimally benefit from a higher number of cores. In this paper, we try to understand the scalability behaviour of network/CPU intensive applications running on multicore architectures. We begin by benchmarking the Petstore web application, noting the systematic imbalance that arises with respect to per-core workload. Having identified the reason for this phenomenon, we propose a queueing model which, when appropriately parametrised, reflects the trend in our benchmark results for up to 8 cores. Key to our approach is providing a fine-grained model which incorporates the idiosyncrasies of the operating system and the multiple CPU cores. Analysis of the model suggests a straightforward way to mitigate the observed bottleneck, which can be practically realised by the deployment of multiple virtual NICs within our VM. Next we make blind predictions to forecast performance with multiple virtual NICs. The validation results show that the model is able to predict the expected performance with relative errors ranging between 8 and 26 per cent.

Performance Evaluation of NoSQL Databases

NoSQL databases have emerged as a backend to support Big Data applications. NoSQL databases are characterized by horizontal scalability, schema-free data models, and easy cloud deployment. To avoid overprovisioning, it is essential to be able to identify the correct number of nodes required for a specific system before deployment. This paper benchmarks and compares three of the most common NoSQL databases: Cassandra, MongoDB and HBase. We deploy them on the Amazon EC2 cloud platform using different types of virtual machines and cluster sizes to study the effect of different configurations. We then compare the behavior of these systems to high-level queueing network models. Our results show that the models are able to capture the main performance characteristics of the studied databases and form the basis for a capacity planning tool for service providers and service users.

Performance modelling of database contention using queueing petri nets

Most performance evaluation studies of database systems are high level studies limited by the expressiveness of their modelling formalisms. In this paper, we illustrate the potential of Queueing Petri Nets as a successor of traditionally-adopted modelling formalisms in evaluating the complexities of database systems. This is demonstrated through the construction and analysis of a Queueing Petri Net model of table-level database locking. We show that this model predicts mean response times better than a corresponding Petri net model.

Modelling Replication in NoSQL Datastores

Distributed NoSQL datastores have been developed to cater for the usage scenarios of Web 2.0 applications. These systems provide high availability through the replication of data across different machines and data centers. The performance characteristics of NoSQL datastores are determined by the degree of data replication and the consistency guarantees required by the application. This paper presents a novel performance study of the Cassandra NoSQL datastore deployed on the Amazon EC2 cloud platform. We show that a queueing Petri net model can scale to represent the characteristics of read workloads for different replication strategies and cluster sizes. We benchmark one Cassandra node and predict response times and throughput for these configurations. We study the relationship between cluster size and consistency guarantees on cluster performance and identify the effect that node capacity and configuration has on the overall performance of the cluster.

Performance Modelling of Concurrency Control Schemes for Relational Databases

The performance of relational database systems is influenced by complex interdependent factors, which makes developing accurate models to evaluate their performance a challenging task. This paper presents a novel case study in which we develop a simple queueing Petri net model of a relational database system. The performance of the database system is evaluated for three different concurrency control schemes and compared to the results predicted by a queueing Petri net model. The results demonstrate the potential of our modelling approach in modelling database systems using relatively simple models that require minimal parameterization. Our models gave accurate approximations of the mean response times for shared and exclusive transactions with average prediction errors of 10% for high contention scenarios.

Teaching Software Engineering in Developing Countries: A Position Paper

This paper presents a new perspective on the issue of localization and contextualization of the undergraduate software engineering curriculum in developing countries. We propose to supplement the international software engineering curriculum with the history, methods, techniques, anecdotes and experiences of the local software industries, its people, successes, failures and environment. The aim is to provide a more realistic student experience that connects international methods and processes with local examples. Thus producing graduates that are familiar with the local challenges and realities that have the capabilities to adapt and transform their acquired skills to meet the local context. This paper is based on the authors experience in teaching software engineering in a developing country and in working within that countrys local software industry.

Approximating closed fork-join queueing networks using product-form stochastic Petri-nets

Product-form SPN approximation for fork-join networks with interfering requests.Conditions for accurate approximation.Validation against simulation.Evaluation of the performance of replication in NoSQL cloud datastores. Computing paradigms have shifted towards highly parallel processing and massive replication of data. This entails the efficient distribution of requests and the synchronization of results provided to users. Guaranteeing SLAs requires the ability to evaluate the performance of such systems while taking the effect of non-parallel workloads into consideration. This can be achieved with performance models that are able to represent both parallel and sequential workloads. This paper presents a product-form stochastic Petri-net approximation of fork-join queueing networks with interfering requests. We derive the necessary conditions that guarantee the accuracy of the approximations and verify this through examples in comparison to simulation. We apply these approximate models to the performance evaluation of replication in NoSQL cloud datastores and illustrate the composition of large models from smaller models, thus facilitating the ability to model a range of deployment scenarios. We show the efficiency of our solution method, which finds the product-form solution of the models without the representation of the state-space of the underlying CTMC.

Benchmarking Replication in Cassandra and MongoDB NoSQL Datastores

The proliferation in Web 2.0 applications has increased the volume, velocity, and variety of data sources which have exceeded the limitations and expected use cases of traditional relational DBMSs. Cloud serving NoSQL data stores address these concerns and provide replication mechanisms to ensure fault tolerance, high availability, and improved scalability. In this paper, we empirically explore the impact of replication on the performance of Cassandra and MongoDB NoSQL datastores. We evaluate the impact of replication in comparison to non-replicated clusters of equal size hosted on a private cloud environment. Our benchmarking experiments are conducted for read and write heavy workloads subject to different access distributions and tunable consistency levels. Our results demonstrate that replication must be taken into consideration in empirical and modelling studies in order to achieve an accurate evaluation of the performance of these datastores.

Software Industry Practice in Africa: Case Study Sudan

The software development industry has a vital role in the economic development of Africa. However, there are limited studies investigating the software industry in Africa. In this work, we attempt to close this gap through the study of the software industry in Sudan (East Africa). This paper presents a questionnaire based survey exploring the demographics, software development practices and risks associated with the Sudanese software industry. Approximately 100 practitioners participated in this study representing the major software companies operating in Sudan. The key challenges identified in this work were the insufficient description of software development procedures, communication methods and customer relations within companies. The results of this study contribute to the diversity of international research in software engineering practice and provide necessary data for researchers in ICT in Africa.