NoSQL databases focus on analytical processing of large scale datasets, offering increased scalability over commodity hardware. One of their strongest features is elasticity, which allows for fairly portioned premiums and high-quality performance and directly applies to the philosophy of a cloudbased platform. Yet, the process of adaptive expansion and contraction of resources usually involves a lot of manual effort during cluster configuration. To date, there exists no comparative study to quantify this cost and measure the efficacy of NoSQL engines that offer this feature over a cloud provider. In this work, we present a cloud-enabled framework for adaptive monitoring of NoSQL systems. We perform a study of the elasticity feature on some of the most popular NoSQL databases over an open-source cloud platform. Based on these measurements, we finally present a prototype implementation of a decision making system that enables automatic elastic operations of any NoSQL engine based on administrator or application-specified constraints.

No abstract

SUMMARYWe investigate the packet-scheduling function within the access scheme of a unidirectional satellite system providing point-to-multipoint services to mobile users. The satellite system may be regarded as an overlay multicast/broadcast layer complementing the point-to-point third generation (3G) mobile terrestrial networks. The satellite access scheme features maximum commonalties with the frequency division duplex (FDD) air interface of the terrestrial universal mobile telecommunications system (T-UMTS), also known as wideband code division multiple access (WCDMA), thus enabling close integration with the terrestrial 3G mobile networks and cost-efficient handset implementations. We draw our attention on one of the radio resource management entities relevant to this interface: the packet scheduler. The lack of channel-state information and the point-to-multipoint service offering differentiate the packet scheduler in the satellite radio interface from its counterpart in point-to-point terrestrial mobile networks. We formulate the scheduler tasks and describe adaptations of two well-known scheduling disciplines, the multilevel priority queuing and weighted fair queuing schemes, as candidates for the time-scheduling function. Simulation results confirm the significance of the transport format combination set (TFCS) with respect to both the resource utilization achieved by the scheduler and the performance obtained by the flows at packet-level. The performance gap of the two schemes regarding the fairness provided to competing flows can be narrowed via appropriate selection of the TFCS, whereas the achieved delay and delay variation scores are ultimately dependent on the packet-level dynamics of individual flows.

SUMMARYThis paper proposes a complete satellite access network solution for multimedia broadcast multicast service (MBMS) delivery based on T-UMTS standards. First, the benefits of MBMS delivery via satellite (SAT-MBMS) for both S/T-UMTS network operators are shown with market and business analysis. A new integrated S/T-UMTS architecture for MBMS delivery is proposed featuring an intermediate module repeater (IMR) for coverage of urban areas. The architectural options of IMR and terminals are discussed considering the relevant cost and complexity. The IMR propagation channel conditions are investigated and a new propagation channel model is proposed. The potential of advanced coding schemes such as the layered coding technique to tackle the channel variations in broadcast/multicast environment is outlined. The functional and protocol architecture are defined along with the interface between the satellite access network and the UMTS core network. Required modifications on the terrestrial access scheme sub-layers to support MBMS data are investigated and the relevant logical, transport and physical channels are selected. Based on the channel selection and the point-to-multipoint service nature, we define a generic radio resource management (RRM) strategy that takes into account both QoS and GoS requirements. The efficiency of the proposed solutions is evaluated in the presented simulation results, advocating the feasibility of the overall approach.

In this paper, we present a distributed architecture for indexing and serving large and diverse datasets. It incorporates and extends the functionality of Hadoop, the open source MapReduce framework, and of HBase, a distributed, sparse, NoSQL database, to create a fully parallel indexing system. Experiments with structured, semi-structured and unstructured data of various sizes demonstrate the flexibility, speed and robustness of our implementation and contrast it with similarly oriented projects. Our 11 node cluster prototype managed to keep full-text indexing time of 150GB raw content in less than 3 hours, whereas the system's response time under sustained query load of more than 1000 queries/sec was kept in the order of milliseconds.

Modern cloud infrastructures based on virtual hardware provide new opportunities and challenges for developers and system administrators alike. Most notable is the promise of resource elasticity, whereby the infrastructure can increase or decrease in size based on demand. Utilizing elastic resources, applications can provide better quality of service and reduce cost by only paying for the required amount of resources. In this work, we extensively study the performance of some popular NoSQL databases over an elastic cloud infrastructure. NoSQL databases focus on analytical processing of large scale datasets, offering increased scalability over commodity hardware. We then proceed to describe TIRA-MOLA, a cloud-enabled framework for automatic provisioning of elastic resources on any NoSQL platform. Our system administers cluster resources (VMs) according to useror application-specified constraints through an expandable monitoring and command-issuing module. Users can easily modify resizing policies, based on application-specific metrics and thus fully utilize the elasticity of the underlying infrastructure. As a realistic use-case, we apply this framework on top of a fully distributed RDF store backed by an elastic NoSQL database. Letting TIRAMOLA manage the number of committed resources results in automated cluster resize actions and throughput maximization, while application experts need only provide simple elasticity rules.

No abstract

No abstract