Optimizing the Content Delivery Network system configuration has been addressed as an interesting problem for the system owners. They want to minimize the investment cost while guaranteeing their system's quality. Several works have resolved this problem as a single-objective optimization (SOO) problem with heuristic methods. These approaches usually aggregate the objectives into a scalar function and resolve the problem with SOO algorithms. A typical drawback of these approaches is that they cannot capture the trade-off between the objectives, which usually leads to a sub-optimal solution. To overcome this drawback, this paper considers the problem as a discrete multiobjective problem and resolves it with meta-heuristic techniques, namely Bayesian optimization (BO) and evolutionary methods. More importantly, we also propose an empirical method to improve the convergence speed of the standard BO methods in discrete space. Our experiments show that our proposed method can dramatically improve the rate of convergence. Moreover, we apply our method to a real CDN system and compare our solution with the system's current solution. Our experimental results show that our proposed solution can save about 39% of the current cost with the same internal traffic.

Analyzing Content Delivery Network (CDN) systems’ performance can help the CDNs’ owners increase their benefits, improve the systems’ service quality, and reduce investment costs. To analyze caching systems, besides mathematical models, there are simulation-based methods, which are based on virtual environments to mimic the system behaviors, typically, such as CDN simulators and testbeds. But these simulators and testbeds have their limitations. In particular, simulators cannot guarantee the functional realism of service applications. Meanwhile, testbeds are cumbersome and inflexible to adapt to new topologies or caching strategies. In this work, we develop a CDN emulator, which is adaptive, flexible, high-fidelity, and reliable. To evaluate the reproducibility of experimental results, we also run our tools multiple configuration parameter sets and compare the results with a discrete event simulation. Our experiments show that our emulator is stable and works properly. The tool’s resource consumption is also evaluated.

Internet traffic is growing quickly, and it is majorly contributed by the proliferation of video services. Content Delivery Networks (CDNs) reduce the video traffic by storing replicas of videos in their cache servers. Nonetheless, the cache servers are usually located outside Internet Service Providers (ISPs). This implies that CDNs cannot reduce the video traffic inside ISP networks. To mitigate this issue, many ISPs build their own CDNs called Telco-CDNs. Genetic Algorithm-based caching is deemed the best approach in terms of traffic reduction. However, it is not practical since its computation time to generate content allocations is extremely long even when using a cluster. A color-based approach was devised to help overcome the drawback at the expense of its increase in traffic. Nevertheless, in case the number of content categories or requests proliferates quickly, the approach also has the same limitation like the Genetic Algorithm-based caching. To resolve the limitation, we propose two novel techniques to hamper the increase in the computation time. One is able to cope with the situation when the number of content categories increases while the other can deal with the circumstance when the number of requests rises. The empirical results show that the computation time is reduced 5x for the former and 7x for the latter at the expense of 1% and 12% increase in traffic for a problem of 5,000 contents, respectively.

With the rapid growth of Internet technology and infrastructure, we have entered the era of data explosion. Following this is the emergence of social networks, which have brought an enormous and ever-growing amount of online content into our digital world. Knowing precisely the popularity of online contents is of great importance for developing advanced caching algorithms as well as content distribution strategies. In this study, we provide some crucial insights into the characteristics of online content popularity over time in different locations and propose a simple predictive model to estimate the popularity of online contents in particular periods. By experiencing with the real datasets of MovieLens and Youtube, our model not only achieves considerable accuracy but also shows an impressive reduction in computation time, from 80 to 250 times faster comparing to some baseline methods. At last, we also provide the potentials and limitations of our model in practice.

Mobile core networks have seen serious network resource contention among a huge amount of delay‐sensitive and delay‐insensitive traffic leading to network congestion or even failures. Delay‐insensitive traffic which is commonly generated by background applications or the operating systems to update the system's status can be postponed to yield network resources for delay‐sensitive traffic, especially in the peak times. This study proposes an efficient approach to delay‐insensitive traffic control leveraging software defined network (SDN) techniques. The study devises an effective data structure, namely flow tree , to resolve the flow aggregation, an essential issue in SDN‐based traffic control, especially when dealing with a huge number of small delay‐insensitive traffic flows. The proposed approach significantly reduces the inherent communication cost between the controller and Openflow switches and the storage cost in switches' expensive memories. This approach is significantly robust in controlling a huge number of small flows which are common in the modern mobile core networks and Internet of things environments. Experimental results from both simulated and real datasets reveal the effectiveness and efficiency of the proposed scheme.

Core networks have to contend with a huge amount of traffic, making the networks heavy, congested or even fail, specifically at peak times. However, not all traffic has to be processed immediately. Background (BG) traffic can be delayed to yield network resources for foreground (FG) traffic when needed. This paper proposes an efficient approach to BG and FG traffic control leveraging software defined network (SDN) techniques. We devised an effective data structure, namely a flow tree, for flow aggregation, to reduce the inherent communication costs (between the Controller and Openflow switches) and the storage cost in the switch's expensive memory. Experimental results from both simulated and real datasets reveal the effectiveness and efficiency of the proposed scheme, confirming the viability of the proposed SDN-based BG and FG traffic control approach.

MPI one-sided communication decouples data movement from synchronization, which eliminates overhead from unneeded synchronization and allows for greater concurrency. On the one hand this fact is the great advantage of MPI one-sided communication, but on the other, it poses enormous challenges for programmers in preserving the reliability of programs. Memory consistency errors are notorious for degrading reliability as well as performance of MPI one-sided applications. Even an MPI expert can easily make these mistakes. The lockopts bug occurred in an RMA test case that is part of MPICH MPI implementation is an example for this situation. Hence, detecting memory consistency errors is extremely challenging. MC-Checker is the most cutting-edge debugger to address these errors effectively. MC-Checker tackles the memory consistency errors based on the happened-before relation. Taking full advantage of the relation makes DN-Analyzer of MC-Checker difficult to scale well. For that reason, MC-Checker does ignore the transitive ordering of the happened-before relation to retain scalability of DN-Analyzer. Consequently, MC-Checker is highly able to impose a potential source of false positives.

Background (BG) traffic generated automatically by applications or the operating system on devices is delay-insensitive traffic as it does not immediately affect user experiences. However, this traffic may significantly contend network resources with foreground (FG) traffic which conveys data/content to users. This contention directly affects the QoS and QoE of network services, specifically at the peak times. This paper proposes an integration approach to which delay-insensitive traffic is quickly detected and effectively transferred in accordance with the network conditions hence the best resources are yielded to FG traffic. For the traffic transferring, we propose two methods namely the router centric (RC) and the SDN-based approaches. In the RC method, the edge routers analyze network conditions, compute the suggested sending rates for individual traffic flows and send these values back to clients where the adjustment mechanisms are implemented. In the SDN-based method the course of actions for each traffic flow is computed by a remote controller. Experimental results reveal the effectiveness and the feasibility of the proposed approaches.

The utilization and flexibility of using resource embrace each organization builds its own private cloud. Cloud Federation is a recent paradigm that helps deal with both physical resources under-utilization and resource limitation in their local infrastructure. Due to the management separation, we are interested in self-sufficiency approach in discovering and matchmaking to establish the federation with other clouds. We leverage event-based design helps communicate over sites to enable negotiate scheduling between among providers. We propose decentralized broker supporting differentiated ratio-base job scheduling (DrbCF) to manage send/receive workload. Our evaluation shows the ability to accept new small capacity member and improve overall federation throughput with the differentiated rate approximately 14-16%. DrbCF is a useful tool supporting the differentiated capacity sites as well as the dynamic of interconnect structure under cloud circumstance.

Environmental issues can be simulated in High Performance Computing (HPC) systems such as simulating flood inundation. This paper presents the simulation of two problems in Vietnam using TELEMAC-MASCARET, namely flood and salinization of Ho Chi Minh city and Mekong Delta. Evaluating these simulations on both physical machines and virtualization platforms are introduced in detail. To meet the requirements of serving multiple users or different simulation problems, virtualization environment offers feasible solutions and ensures the availability as well as the flexibility. Besides the popularity of virtual machine (VM) in Cloud, recently there is a lightweight virtualization platform called Docker. In this paper, we propose a model for deploying TELEMAC on Docker. Then, we show evaluations through different scenarios when deploying TELEMAC on Docker and VM. Docker has more potential with the lightweight architecture, especially its overhead is nearly negligible. The performance of Docker is better than VM, and the execution time of TELEMAC can be reduced significantly. The execution time of VM is about 1.6 times longer than Docker in terms of running the simulations of environment-related problems in Vietnam.

Debugging large-scale parallel applications is a problematic issue. Characteristics of scalability bring about an exponential increase in errors and many impacts on performance. With suffering unacceptable overhead and debugging time, traditional techniques, such as checkpointing or record and replay, have become obsolete when applying to largescale parallel applications. The ex-scale trend is coming, which demands cutting-edge large-scale parallel application debugging techniques. Instead of prior works based on locating exact errors, we proposed an on-the-fly approach by detecting abnormal behaviors arising frequently in complicated message passing channels. In this paper, anomalies are race conditions causing concealing deadlocks which probably result in hangs and make programmers unable to inspect manually errors. The technique utilizes one state-of-the-art detection algorithm which is related to allocation and management tactics. The proposed algorithm is proved the precision and effectiveness by theoretical proofs and experimental results. With acceptable overhead, this technique shows the potential for applying to large-scale parallel applications, specially ones running as master/slave model.

Many Integrated Core (MIC) architecture systems are becoming increasingly popular for HPC applications as they have the dual-advantage of accelerating vector processing and a general-purpose programming model. One of the key challenges for energy-efficient execution on MIC architecture systems is to determine time and energy-efficient configurations among a large system configuration space. Given a parallel program with an execution time deadline and an energy budget, we propose a measurement-based analytical modeling approach to determine these system configurations. In contrast to current approaches, we model both inter-and intra-core resource overlaps, memory contention among threads within a core and memory contention across multiple cores. The model is validated against direct measurement using six representative HPC applications on Intel Xeon Phi system. We show that a Pareto frontier consisting of optimal configurations exist for a parallel program executing on MIC architecture systems. To further understand the Pareto frontier, we use the performance-to-power ratio metric (PPR), that quantifies the amount of useful computations performed per unit energy during the execution. The proposed approach can be used to determine what thread affinity is suitable for offloading execution to accelerators such as Xeon Phi and save energy.

Recent growth of traffic surveillance based on computer vision techniques has caught more and more attentions from researchers. Since the detection of vehicles is the primary step of such system, there is a large body of works towards developing an efficient detection scheme on various operating conditions. However, those works mainly focus on the detection of car and pedestrian. In this paper, we shift our attention to motorbike, which is also a common road user, especially in developing country. In comparison to other target objects, motorbike is rather small in size but has more complex structure. Thus, detecting motorbike is not a trivial task in real-life context where there is variance due to presence of motorbike drivers and high degree of occlusions. To address this problem, we propose a method for detecting motorbike from the scenes. Our method can achieve robustness to changes in illuminations, affine transformations and occlusions. Firstly, local features are extracted from images which contain one single target object. These local features represent parts of objects and are used to construct a Bag-of-Visual Words model. Using this model, each object is represented as a histogram of their parts. Next, a Support Vector Machine classifier is trained with these representations for classifying motorbike and non-motorbike objects. Finally, we develop an algorithm to form a detection hypothesis to detect multiple target objects from the scene. We collect a dataset of 3000 images for evaluating our proposed method. The experimental results indicate that our method can achieve high accuracy in the context of real-life motorbike detection applications.

Application layer multicast (ALM) is an alternative to overcome many problems that prevent IP multicast from widely using. However, when the overlay route changes, it takes time for end-hosts to update new routing table. In this paper, we propose Treemap a technique to alleviate this problem. The main contribution of Treemap is to perform application layer routing which is self contained in a packet. As a result, end-hosts do not need to update new overlay route, yielding fast route adaptation when the overlay route changes. Furthermore, the Treemap technique can be used to extend Xcast6 (Explicit Multi-unicast for IPv6). Xcast6 Treemap sends data in network layer multicast with the support of Xcast router, otherwise it works like ALM. With nominal packet header processing, Xcast6 Treemap is well suited for loss/time sensitive applications such as video conference.

In this paper we propose bandwidth fair N-Tree algorithm for ALM distribution tree construction and a new protocol for ALM packet replication and distribution, namely Almcast. Both the tree construction algorithm and packet replication/distribution protocol were implemented as proof-of concept by modifying an existing multi-party video conference application. The results show that N-tree algorithm takes less than 3 ms to construct ALM distribution tree for 12 nodes. Almcast implementation enables the intermediate relay node to lookup for next destination, replicate and forward packets as fast as its receiving rate at application layer.

This paper presents a checkpoint and recovery (C&R) protocol to support fault-tolerance for PVM (Parallel Virtual Machine). The protocol helps to mask fail-stop failures from an application. The C&R activities are transparent and do not require any change in the PVM library nor operating system. In PVM, an application can change the number of processes during execution. This paper focuses on solving problems raised by the dynamic spawn and the asynchronous exit of tasks in PVM. The proposed protocol is a non-blocking one, so it reduces side-effect of checkpoint activities of original programs.