Search

peer-reviewed Synchronisation mechanisms are essential in distributed sim- ulation. Some systems rely on central units to control the simulation but central units are known to be bottlenecks [10]. If we want to avoid using a central unit to optimise the simulation speed, we lose the capacity to act on the sim- ulation at a global scale. Being able to act on the entire simulation is an important feature which allows to dynam- ically load-balance a distributed simulation. While some local partitioning algorithms exist [12], their lack of global view reduces their e ciency. Running a global partitioning algorithm without central unit requires a synchronisation of all logical processes (LPs) at the same step.We introduce in this paper two algorithms allowing to synchronise logical processes in a distributed simulation without any central unit. The rst algorithm requires the knowledge of some topological properties of the network while the second algo- rithm works without any requirement. The algorithms are detailed and compared against each other. An evaluation shows the bene ts of using a global dynamic load-balancing for distributed simulations.

Using modelling to predict the performance characteristics of software applications typically uses Queueing Network Models representing the various system hardware resources. Leaving out the software resources, such as the limited number of threads, in such models leads to a reduced prediction accuracy. Accounting for Software Contention is a challenging task as existing techniques to model software components are complex and require deep knowledge of the software architecture. Furthermore, they also require complex measurement processes to obtain the model's service demands. In addition, solving the resultant model usually require simulation solvers which are often time consuming.In this work, we aim to provide a simpler model for three-tier web software systems which accounts for Software Contention that can be solved by time efficient analytical solvers. We achieve this by expanding the existing "Two-Level Iterative Queuing Modelling of Software Contention" method to handle the number of threads at the Application Server tier and the number of Data Sources at the Database Server tier. This is done in a generic manner to allow for extending the solution to other software components like memory and critical sections. Initial results show that our technique clearly outperforms existing techniques.

peer-reviewed Computing servers generally have a narrow dynamic power range. For instance, even completely idle servers consume between 50% and 70% of their peak power. Since the us- age rate of the server has the main in uence on its power consumption, energy-e ciency is achieved whenever the uti- lization of the servers that are powered on reaches its peak. For this purpose, enterprises generally adopt the following technique: consolidate as many workloads as possible via virtualization in a minimum amount of servers (i.e. maxi- mize utilization) and power down the ones that remain idle (i.e. reduce power consumption). However, such approach can severely impact servers' performance and reliability. In this paper, we propose a methodology to determine the ideal values for power consumption and utilization for a server without performance degradation. We accomplish this through a series of experiments using two typical types of workloads commonly found in enterprises: TPC-H and SPECpower ssj2008 benchmarks. We use the rst to mea- sure the amount of queries responded successfully per hour for di erent numbers of users (i.e. Throughput@Size) in the VM. Moreover, we use the latter to measure the power con- sumption and number of operations successfully handled by a VM at di erent target loads. We conducted experiments varying the utilization level and number of users for di er- ent VMs and the results show that it is possible to reach the maximum value of power consumption for a server, without experiencing performance degradations when running indi- vidual, or mixing workloads.

The emerging Media Independent Handover (MIH) standard proposes to support session continuity during handover between heterogeneous networks. One of the critical features provided by MIH is an Event Service which includes predictive network degradation events, such as Link_Going_Down (LGD), which are triggered based on link layer metrics. Our results highlight the reactivness of media stream quality to network degradation. The point of degradation however, is specific to the characteristics of the class of media streaming service. Many existing event algorithms utilize static performance thresholds which are unresponsive to the requirements of individual application service classes. In this paper we propose a Service Oriented Link Triggering Algorithm (SOLTA) which triggers the LGD and Link_Down (LD) events based on link layer metrics but subject to the performance characteristics of the supported class of service. SOLTA illustrates that for 802.11, it is necessary to have aggressive service class specific, link event triggering. SOLTA also illustrates how a soft path handover approach such as Stream Control Transmission Protocols Concurrent Multi-path Transfer (SCTP-CMT) variant is necessary to support seamless session migration.

In this paper, an interference aware algorithm for multi-cell Relay Station (RS) location planning problem is proposed in order to investigate the capacity gains possible using transparent mode 802.16j. The study focuses on the problem of determining the best choice of locations for RS from a given set of candidate sites - Base Station (BS) and Subscriber Station (SS) locations are assumed given. The proposed algorithm evaluates each RS based on the channel conditions of the links between the RS and its surrounding Subscriber Stations(SS) and the throughput gain it could deliver, taking into account co-channel interference. The performance statistics were collected over a large amount of experiments with many different configurations, including Base Station(BS) antenna types, number of candidate RS sites, RS transmitting power, with/without spatial reuse, etc. The results can be divided into two sections. Firstly, the algorithm was tested for optimality and scalability and was shown to perform well compared to classical Integer Programming solution techniques. Secondly, the results generated by the algorithm when applied to different scenarios were analyzed. Results show that using spatial reuse could achieve a much higher throughput gain, 3-5 time, then without using spatial reuse, although sectorised BS antenna limits the separation of co-channel RSs in a sector. The results demonstrate the characteristics of the trade-off between number of RSs and throughput gain for different configurations.

The existence of hidden and exposed nodes can have a significantly negative impact on the performance of IEEE802.11 networks. Such nodes can increase the probability of collisions and limit the spatial reuse on the channel. The value of the Physical Carrier Sensing Threshold (PCST) is instrumental to the trade-off between the number of hidden and exposed nodes in a system.This paper presents a new adaptive physical carrier sensing mechanism for wireless networks. The K-APCS algorithm maximizes the aggregate system throughput by optimizing the balance between the spatial reuse and the collision rate of a system. It incorporates 802.11k radio resource measurements and a frame loss bound on each node to enable the on-line tuning of the PCST. Simulations were performed on five variants of carrier sensing mechanisms; results show that K-APCS achieves the maximum spatial reuse with a throughput gain of 28% and the minimum collision rate with a decrease of 0.01%.

We present a transparent optimisation framework for automatically improving run-time performance of component-based enterprise applications. Run-time performance is improved by automatically identifying and dynamically switching to an optimised but functionally equivalent program flow, for a specific transaction type. More precisely, one or more application components can have both static and dynamic inputs, where the former and its output remain mostly fixed from invocation-to-invocation, while the latter is unconstrained. Through dynamic profiling and run-time adaptation, an optimised program flow is derived, that allows only future execution of program points that operate on dynamic inputs and not static inputs. During execution, a target application is instrumented to switch to an optimised program flow. Additional instrumentation occurs to profile key program points that may invalidate an optimised program flow. If these points execute, then the application reverts back to its pre-optimised program flow. We evaluate and demonstrate how run-time performance for a typical thre-tired enterprise system can be improved for frequently accessed data that remains mostly static. For the application studied we showed a 49% marked improvement in the number of users serviced per second.

Nowadays component technologies are an integral part of any enterprise production environment. Performance and scalability are among the key properties of such systems. Using Layered Queuing Networks (LQN), one can predict the performance of a component based system from its design. This work revises the approach of using LQN templates, and offers a case study by using the revised approach to model a realistic component application.

This work describes a detailed simulation-based study of the performance of an IEEE 802.11e Medium Access Control (MAC) layer over an IEEE 802.11g Physical (PHY) layer. The study focuses on the number of simultaneous bidirectional G.711 Voice over IP (VoIP) calls that can be supported by a Wireless Local Area Network (WLAN) using the Extended Rate PHY - Orthogonal Frequency Division Modulation (ERP-OFDM) mode of 802.11g.A new scheme for adapting the WLAN MAC parameters is proposed in this work. The new scheme dynamically adapts the Contention Window (CW) based on the retransmission rate of the system. In addition, an adaptive Transmission Opportunity (TXOP) mechanism aids in balancing the uplink and downlink traffic levels and so provides the equality in uplink and downlink performance that is required for bidirectional VoIP traffic. The proposed scheme can thus maintain acceptable levels of QoS for higher call capacities, increasing the overall VoIP capacity of the system.