Nam Thoai

A genetic algorithm for power-aware virtual machine allocation in private cloud

Energy efficiency has become an important measurement of scheduling algorithm for private cloud. The challenge is trade-off between minimizing of energy consumption and satisfying Quality of Service (QoS) (e.g. performance or resource availability on time for reservation request). We consider resource needs in context of a private cloud system to provide resources for applications in teaching and researching. In which users request computing resources for laboratory classes at start times and non-interrupted duration in some hours in prior. Many previous works are based on migrating techniques to move online virtual machines (VMs) from low utilization hosts and turn these hosts off to reduce energy consumption. However, the techniques for migration of VMs could not use in our case. In this paper, a genetic algorithm for power-aware in scheduling of resource allocation (GAPA) has been proposed to solve the static virtual machine allocation problem (SVMAP). Due to limited resources (i.e. memory) for executing simulation, we created a workload that contains a sample of one-day timetable of lab hours in our university. We evaluate the GAPA and a baseline scheduling algorithm (BFD), which sorts list of virtual machines in start time (i.e. earliest start time first) and using best-fit decreasing (i.e. least increased power consumption) algorithm, for solving the same SVMAP. As a result, the GAPA algorithm obtains total energy consumption is lower than the baseline algorithm on simulated experimentation.

EPOBF: ENERGY EFFICIENT ALLOCATION OF VIRTUAL MACHINES IN HIGH PERFORMANCE COMPUTING CLOUD

Cloud computing has become more popular in provision of computing resources under virtual machine (VM) abstraction for high performance computing (HPC) users to run their applications. A HPC cloud is such cloud computing environment. One of challenges of energy efficient resource allocation for VMs in HPC cloud is tradeoff between minimizing total energy consumption of physical machines (PMs) and satisfying Quality of Service (e.g. performance). On one hand, cloud providers want to maximize their profit by reducing the power cost (e.g. using the smallest number of running PMs). On the other hand, cloud customers (users) want highest performance for their applications. In this paper, we focus on the scenario that scheduler does not know global information about user jobs and user applications in the future. Users will request shortterm resources at fixed start times and non interrupted durations. We then propose a new allocation heuristic (named Energy-aware and Performance per watt oriented Bestfit (EPOBF)) that uses metric of performance per watt to choose which most energy-efficient PM for mapping each VM (e.g. maximum of MIPS per Watt). Using information from Feitelsons Parallel Workload Archive to model HPC jobs, we compare the proposed EPOBF to state of the art heuristics on heterogeneous PMs (each PM has multicore CPU). Simulations show that the EPOBF can reduce significant total energy consumption in comparison with state of the art allocation heuristics.

Future Data and Security Engineering

The amount of unstructured data has grown exponentially during the past two decades and continues to grow at even faster rates. As a consequence, the efficient management of this kind of data came to play an important role in almost all organizations. Up to now, approaches from many different research fields, like information search and retrieval, text mining or query expansion and reformulation, have enabled us to extract and learn patterns in order to improve the management, retrieval and recommendation of documents. However, there are still many open questions, limitations and vulnerabilities that need to be addressed. This paper aims at identifying the current major challenges and research gaps in the field of “document enrichment, retrieval and recommendation”, introduces innovative ideas towards overcoming these limitations and weaknesses, and shows the benefits of adopting these ideas into real enterprise content management systems.

Using Docker in high performance computing applications

Virtualization technology plays a vital role in cloud computing. In particular, benefits of virtualization are widely employed in high performance computing (HPC) applications. Recently, virtual machines (VMs) and Docker containers known as two virtualization platforms need to be explored for developing applications efficiently. We target a model for deploying distributed applications on Docker containers, among using well-known benchmarks to evaluate performance between VMs and containers. Based on their architecture, we propose benchmark scenarios to analyze the computing performance and the ability of data access on HPC system. Remarkably, Docker container has more advantages than virtual machine in terms of data intensive application and computing ability, especially the overhead of Docker is trivial. However, Docker architecture has some drawbacks in resource management. Our experiment and evaluation show how to deploy efficiently high performance computing applications on Docker containers and VMs.

Heuristics for Energy-Aware VM Allocation in HPC Clouds

High performance computing (HPC) clouds have become more popular for users to run their HPC applications on cloud infrastructures. Reduction in energy consumption (kWh) for these cloud systems is of high priority for any cloud provider. In this paper, we first study the energy-aware allocation of virtual machines (VMs) in HPC cloud systems along two dimensions: multi-dimensional resources and interval times of virtual machines. On the one hand, we present an example showing that using bin-packing heuristics (e.g. Best-Fit Decreasing) to minimize the number of physical servers could not lead to a minimum of total energy consumption. On the other hand, we find out that minimizing total energy consumption is equivalent to minimizing the sum of total completion time of all physical machines. Based on this finding, we propose the MinDFT-ST and MinDFT-FT algorithms to place the VMs onto the physical servers in such a way that minimizes the total completion times of all physical servers. Our simulation results show that MinDFT-ST and MinDFT-FT could reduce the total energy consumption by 22.4% and respectively 16.0% compared with state-of-the-art power-aware heuristics (such as power-aware best-fit decreasing) and vector bin-packing norm-based greedy algorithms (such as VBP-Norm-L1, VBP-Norm-L2, VBP-Norm-L30).

Bandwidth Fair Application Layer Multicast for Multi-Party Video Conference Application

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 3ms 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.

Xcast6 treemap islands: revisiting multicast model

Due to the complexity and poor scalability, IP Multicast has not been used on the Internet. Recently, Xcast6 -- a complementary protocol of IP Multicast has been proposed. However, the key limitation of Xcast6 is that it only supports small multicast sessions. To overcome this, we propose Xcast6 Treemap islands (X6Ti) -- a hybrid model of Overlay Multicast and Xcast6. In summary, X6Ti has many advantages: support large multicast groups, simple and easy to deploy on the Internet, no router configuration, no restriction on the number of groups, no multicast routing protocol and no group management protocol. Based on simulation, we compare X6Ti with IP Multicast and NICE protocols to show the benefits of our new model.

Provision of Docker and InfiniBand in High Performance Computing

High Performance Computing (HPC) is playing an important role in a variety of domains with the demand of high-level computational capacity. Besides, HPC provides services for a huge range of different users as well as multiple environments. Hence, the performance of the network is also one of the important criteria. The advent of InfiniBand (IB) aims at the improvement of computer-networking. IB technology has been used and expanded on virtualization environments, especially in virtual machines (VMs). Recently, another virtualization technique known as Docker platform is being popularly considered. Docker promises to bring higher performance, but it also poses some challenging problems. Concretely, while VMs are combined with IB by standard virtualization modules such as SR-IOV, Docker containers are still being examined on feasible solutions with IB. An important question is the advantages and disadvantages of both architectures, namely VM and Docker. In this paper, we deploy Docker on IB infrastructure and evaluate their performance with VMs. Remarkably, we highlight the benefits and the drawbacks of Docker in the conflict of resources when its architecture shares the same kernel with the host. Our evaluations emphasize the potential of Docker containers in HPC field, simultaneously, we propose experiences when using Docker for running parallel applications.

EMinRET: Heuristic for Energy-Aware VM Placement with Fixed Intervals and Non-preemption

Infrastructure-as-a-Service (IaaS) clouds have become more popular enabling users to run applications under virtual machines. This paper investigates the energy-aware virtual machine (VM) allocation problems in IaaS clouds along characteristics: multiple resources, and fixed interval times and non-preemption of virtual machines. Many previous works proposed to use a minimum number of physical machines, however, this is not necessarily a good solution to minimize total energy consumption in the VM placement with multiple resources, fixed interval times and non-preemption. We observed that minimizing total energy consumption of physical machines is equivalent to minimize the sum of total completion time of all physical machines. Based on the observation, we propose EMinRET algorithm. The EMinRET algorithm swaps an allocating VM with a suitable overlapped VM, which is of the same VM type and is allocated on the same physical machine, to minimize total completion time of all physical machines. The EMinRET uses resource utilization during executing time period of a physical machine as the evaluation metric, and will then choose a host that minimizes the metric to allocate a new VM. In addition, this work studies some heuristics for sorting the list of virtual machines (e.g., sorting by the earliest starting time, or the longest duration time first, etc.) to allocate VM. Using the realistic log-trace in the Feitelsons Parallel Workloads Archive, our simulation results show that the EMinRET algorithm could reduce from 25% to 45% energy consumption compared with power-aware best-fit decreasing (PABFD) [1]) and vector binpacking norm-based greedy algorithms (VBP-Norm-L1/L2 [2]). Moreover, the EMinRET heuristic has also less total energy consumption than our previous heuristics (e.g. MinDFT and EPOBF) in the simulations (using same virtual machines sorting method).

A GPU-Based Enhanced Genetic Algorithm for Power-Aware Task Scheduling Problem in HPC Cloud

In this paper, we consider power-aware task scheduling (PATS) in HPC clouds. Users request virtual machines (VMs) to execute their tasks. Each task is executed on one single VM, and requires a fixed number of cores (i.e., processors), computing power (million instructions per second - MIPS) of each core, a fixed start time and non-preemption in a duration. Each physical machine has maximum capacity resources on processors (cores); each core has limited computing power. The energy consumption of each placement is measured for cost calculating purposes. The power consumption of a physical machine is in a linear relationship with its CPU utilization. We want to minimize the total energy consumption of the placements of tasks. We propose here a genetic algorithm (GA) to solve the PATS problem. The GA is developed with two versions: (1) BKGPUGA, which is an adaptively implemented using NVIDIA’s Compute Unified Device Architecture (CUDA) framework; and (2) SGA, which is a serial GA version on CPU. The experimental results show the BKGPUGA program that executed on a single NVIDIA® TESLATM M2090 GPU (512 cores) card obtains significant speedups in comparing to the SGA program executing on Intel® XeonTM E5-2630 (2.3 GHz) on same input problem size. Both versions share the same GA’s parameters (e.g. number of generations, crossover and mutation probability, etc.) and a relative small (10-11) on difference of two finesses between BKGPUGA and SGA. Moreover, the proposed BKGPUGA program can handle large-scale task scheduling problems with scalable speedup under limitations of GPU device (e.g. GPU’s device memory, number of GPU cores, etc.).