A heuristic flow-decomposition approach for generalized processor sharing under self-similar traffic

The well-known Generalized Processor Sharing (GPS) scheduling principle and its variants have received tremendous research efforts due to their appealing properties of fairness, traffic isolation, and work conservation. Traffic self-similarity is highly detrimental to the performance of scheduling mechanisms and communication networks. This paper proposes a novel and heuristic flow-decomposition approach to performance modeling of the GPS system under self-similar traffic. Based on the comprehensive analysis of the excess service sharing behavior of traffic flows, we decompose the GPS system equivalently into a group of single-server single-queue systems. Extensive simulation experiments are used to validate the correctness of the proposed flow-decomposition approach as well as the analytical performance results.     

The multiple phase service network with generalized processor sharing

Summary An analysis is given of multiple phase service facilities of which queueing networks are special models, for the case of a service discipline to be denoted as generalized processor sharing. Under this discipline requests are served simultaneously with a rate depending on the phase and the number of requests present here. The model is of a very general type, its analysis is given for arbitrary routing matrices and absolutely continuous required service time distributions. The mathematical technique used is that of the supplementary variable. Generalisations of known results for closed and open networks are obtained and new results about the average sojourn time of a request in the system are derived, in particular for requests with given route and given processing times at the nodes of the route. Some basic results about reversed processes and departure processes are discussed. For a special but important model the workload is discussed.     

The AMD Opteron processor for multiprocessor servers

Representing AMD's entry into 64-bit computing, Opteron combines the backwards compatibility of the X86-64 architecture with a DDR memory controller and hypertransport links to deliver server-class performance. These features also make Opteron a flexible, modular, and easily connectable component for various multiprocessor configurations.     

Adaptive load sharing for clustered digital library servers

This paper investigates load balancing strategies for clustered Alexandria Digital Library (ADL) servers. The ADL system, which provides online information searching and browsing of spatially-referenced materials through the World Wide Web, involves intensive database I/O and heterogeneous CPU activities. Clustering servers can improve the scalability of the ADL system in response to a large number of simultaneous access requests. One difficulty addressed is that clustered workstation nodes may be non-uniform in terms of CPU and I/O speeds. An optimization scheme is proposed in this paper to dynamically monitor the resource availability, use a low-cost communication strategy for updating load information among nodes, and schedule requests based on both I/O and computation load indices. Since the accurate cost estimation for processing database-searching requests is difficult, a sampling and prediction scheme is used to identify the relative efficiency of nodes for satisfying I/O and CPU demands of these requests. A set of experiments using the ADL traces have been conducted to verify the effectiveness of the proposed strategies.     

Adaptive global power optimization for Web servers

This work investigates power and performance trade-offs for Web servers on a state-of-the-art, high-density, power-efficient SeaMicro SM15k cluster by AMD. We relied on the concept of virtual power states (VPSs), a combination of CPU utilization rate to the P/C power states available in modern processors, and on our global optimization algorithm called Slack Recovery, to deploy an adaptive global power management system in a production environment. The main contributions of this paper are twofold. First, it presents the Slack Recovery algorithm deployed on a real cluster, composed of 25 SeaMicro nodes. The algorithm finds a P-state and a utilization rate for each CPU node to minimize power under a minimum performance requirement. Second, it proposes a novel mechanism to control utilization rates in each server, a key aspect on our power/performance optimization system which enables the implementation of the VPS concept in practice. Experimental results show that our Slack Recovery-based system can reduce up to 6.7 % of the power consumption when compared to policies usually deployed in SeaMicro production systems.     

Hierarchical attribute-based encryption and scalable user revocation for sharing data in cloud servers

With rapid development of cloud computing, more and more enterprises will outsource their sensitive data for sharing in a cloud. To keep the shared data confidential against untrusted cloud service providers (CSPs), a natural way is to store only the encrypted data in a cloud. The key problems of this approach include establishing access control for the encrypted data, and revoking the access rights from users when they are no longer authorized to access the encrypted data. This paper aims to solve both problems. First, we propose a hierarchical attribute-based encryption scheme (HABE) by combining a hierarchical identity-based encryption (HIBE) system and a ciphertext-policy attribute-based encryption (CP-ABE) system, so as to provide not only fine-grained access control, but also full delegation and high performance. Then, we propose a scalable revocation scheme by applying proxy re-encryption (PRE) and lazy re-encryption (LRE) to the HABE scheme, so as to efficiently revoke access rights from users.     

Adaptive piggybacking: a novel technique for data sharing in video-on-demand storage servers

Recent technology advances have made multimedia on-demand services, such as home entertainment and home-shopping, important to the consumer market. One of the most challenging aspects of this type of service is providing access either instantaneously or within a small and reasonable latency upon request. We consider improvements in the performance of multimedia storage servers through data sharing between requests for popular objects, assuming that the I/O bandwidth is the critical resource in the system. We discuss a novel approach to data sharing, termed adaptive piggybacking, which can be used to reduce the aggregate I/O demand on the multimedia storage server and thus reduce latency for servicing new requests.     

Regression-based resource provisioning for session slowdown guarantee in multi-tier Internet servers

Autonomous management of a multi-tier Internet service involves two critical and challenging tasks, one understanding its dynamic behaviors when subjected to dynamic workloads and second adaptive management of its resources to achieve performance guarantees. We propose a statistical machine learning based approach to achieve session slowdown guarantees of a multi-tier Internet service. Session slowdown is the relative ratio of a session’s total queueing delay to its total processing time. It is a compelling performance metric of session-based online transactions because it directly measures user-perceived relative performance and it is independent of the session length. However, there is no analytical model for session slowdown on multi-tier servers. We first conduct training to learn the statistical regression models that quantitatively capture an Internet service’s dynamic behaviors as relationships between various service parameters. Then, we propose a dynamic resource provisioning approach that utilizes the learned regression models to efficiently achieve session slowdown guarantee under dynamic workloads. The approach is based on the combination of offline training and online monitoring of the Internet service behavior. Simulations using the industry standard TPC-W benchmark demonstrate the effectiveness and efficiency of the regression based resource provisioning approach for session slowdown oriented performance guarantee of a multi-tier e-commerce application.     

A generalized processor mapping technique for array redistribution

In many scientific applications, array redistribution is usually required to enhance data locality and reduce remote memory access in many parallel programs on distributed memory multicomputers. Since the redistribution is performed at runtime, there is a performance trade-off between the efficiency of the new data decomposition for a subsequent phase of an algorithm and the cost of redistributing data among processors. In this paper, we present a generalized processor mapping technique to minimize the amount of data exchange for BLOCK-CYCLIC(kr) to BLOCK-CYCLIC(r) array redistribution and vice versa. The main idea of the generalized processor mapping technique is first to develop mapping functions for computing a new rank of each destination processor. Based on the mapping functions, a new logical sequence of destination processors can be derived. The new logical processor sequence is then used to minimize the amount of data exchange in a redistribution. The generalized processor mapping technique can handle array redistribution with arbitrary source and destination processor sets and can be applied to multidimensional array redistribution. We present a theoretical model to analyze the performance improvement of the generalized processor mapping technique. To evaluate the performance of the proposed technique, we have implemented the generalized processor mapping technique on an IBM SP2 parallel machine. The experimental results show that the generalized processor mapping technique can provide performance improvement over a wide range of redistribution problems     

Lazy structure sharing for query optimization

We studylazy structure sharing as a tool for optimizing equivalence testing on complex data types. We investigate a number of strategies for implementing lazy structure sharing and provide upper and lower bounds on their performance (how quickly they effect ideal configurations of our data structure). In most cases when the strategies are applied to a restricted case of the problem, the bounds provide nontrivial improvements over the naïve linear-time equivalence-testing strategy that employs no optimization. Only one strategy, however, which employs path compression, seems promising for the most general case of the problem.Work completed while at Princeton University and supported by a Fannie and John Hertz Foundation Fellowship, National Science Foundation Grant No. CCR-8920505, and the Center for Discrete Mathematics and Theoretical Computer Science (DIMACS) under NSF-STC-91-19999.Work completed while at Princeton University and DIMACS and supported by DIMACS under NSF-STC-91-19999.Research at Princeton University partially supported by the National Science Foundation, Grant No. CCR-8920505, the Office of Naval Research, Contract No. N00014-91-J-1463, and by DIMACS under NSF-STC-91-19999.     

Bandwidth allocation algorithms for weighted maximum rate constrained link sharing policy

This paper addresses the problem of bandwidth allocation under the weighted maximum rate constrained link sharing policy and proves a key theory in the condition of allocation termination. We propose several algorithms with various worst-case and average-case time complexities, and evaluate their computation elapse times.     

Design and implementation of an operating system for composable processor sharing

Multi-Processor Systems on Chip (MPSoC) run multiple independent applications, often developed by different parties. The applications share the hardware resources, e.g. processors, memories and interconnect. The sharing typically causes interference between the applications, which severely complicates system integration and verification. Even if the applications are verified in isolation, the system designer must verify the combined behaviour, leading to an explosion in design complexity. Composable MPSoCs have no interference between applications, thus allowing independent design and verification. For an MPSoC to be composable, all the hardware resources must offer composability. A particularly challenging resource is the processors, often purchased as off-the-shelf intellectual property.In this work we present the design and implementation of CompOSe, a light-weight (only 1500 lines of code) composable operating system for MPSoCs. CompOSe uses fixed-size time slices, coupled with a composable scheduler, to enable composable processor sharing. Using instances of ARM7, ARM11 and the Xilinx MicroBlaze we experimentally demonstrate the ability to provide temporal composability, even in the presence of dynamic application behaviour and multiple use cases. We do so using a diverse set of processor architectures, without requiring any hardware modifications. We also show how CompOSe allows slack to be distributed within and between applications through a novel two-level scheduler and slack-distribution system.     

Queueing models for a single server LAN with processor sharing disciplines

A local area network (LAN) is a collection of autonomous workstations interconnected by a communication network. A key component of a local network is the file server which stores programs and data and makes them available to the workstations as needed. In practice, a workstation requests a large portion of the file (type 1 request) when a new application is launched. Following this, the workstation requests additional portions of the file as needed (type 2 requests). Clearly, the response time to these requests will depend strongly on the file server scheduling policy to service the two types of incoming requests. In an earlier paper [12] we studied this system when type 2 requests havepreemptive and non-preemptive priority over type 1 requests. From a practical point of view, it is more realistic to describe the file transfer by around-robin scheduling discipline. In this paper we study LANs where the file server followsprocessor sharingdiscipline, which is a limiting case of theround-robin discipline. Assuming that all relevant intervals are exponentially distributed we develop algorithms to analyze the performance of the system. Illustrative examples are presented to study the system behavior under various load conditions. The computational approach presented in this paper helps in resolving some of the analytical difficulties associated with the analysis ofprocessor sharing disciplines.     

Parallel extremal optimization in processor load balancing for distributed applications

The paper concerns parallel methods for extremal optimization (EO) applied in processor load balancing in execution of distributed programs. In these methods EO algorithms detect an optimized strategy of tasks migration leading to reduction of program execution time. We use an improved EO algorithm with guided state changes (EO-GS) that provides parallel search for next solution state during solution improvement based on some knowledge of the problem. The search is based on two-step stochastic selection using two fitness functions which account for computation and communication assessment of migration targets. Based on the improved EO-GS approach we propose and evaluate several versions of the parallelization methods of EO algorithms in the context of processor load balancing. Some of them use the crossover operation known in genetic algorithms. The quality of the proposed algorithms is evaluated by experiments with simulated load balancing in execution of distributed programs represented as macro data flow graphs. Load balancing based on so parallelized improved EO provides better convergence of the algorithm, smaller number of task migrations to be done and reduced execution time of applications.     

Throughput optimization for video streaming proxy servers based on video staging

A video streaming proxy server needs to handle hundreds of simultaneous connections between media servers and clients. Inside, every video arrived at the server and delivered from it follows a specific arrival and delivery schedule. While arrival schedules compete for incoming network bandwidth, delivery schedules compete for outgoing network bandwidth. As a result, a proxy server has to provide sufficient buffer and disk cache for storage, together with memory space, disk space and disk bandwidth. In order to optimize the throughput, a proxy server has to govern the usage of these resources. In this paper, we first analyze the property of a traditional smoothing algorithm and a video staging algorithm. Then we develop, based on the smoothing algorithm, a video staging algorithm for video streaming proxy servers. This algorithm allows us to devise an arrival schedule based on the delivery schedule. Under this arrival and delivery schedule pair, we can achieve a better resource utilization rate gracefully between different parameter sets. It is also interesting to note that the usage of the resources such as network bandwidth, disk bandwidth and memory space becomes interchangeable. It provides the basis for inter-resource scheduling to further improve the throughput of a video streaming proxy server system.

Internet瓶颈链路带宽测量方法

网络测量是了解网络行为,进行网络控制,提高网络性能的重要环节和前提基础,其中,瓶颈链路带宽测量一直是网络测量的研究热点之一。该文分别讨论了基于可变分组大小技术和基于分组对(Packetpair)技术的网络瓶颈链路带宽测量算法,首先分析了两类算法的基本原理,然后仔细研究了算法的性能和存在问题,最后提出了进一步的研究方向。     

时频聚集性能优化的广义S变换

S变换是一种具有短时傅里叶变换和小波变换优点的时频分析方法,已有的一些基于能量聚集度量的优化方法集中度不高,影响了信号检测等应用中时频域局部定位的精度.为了提高S变换的时频聚集性能,提出了一种新的时频能量聚集度量方法优化广义S变换,提高了算法的时频集中度和短时傅里叶变换、S变换、广义S变换等方法实验比较,表明提出的方法能有效地提高广义S变换的能量集中度,并具有较强的抗噪声性能.     

On differential equations arising in the theory of processor sharing queues

We give a probabilistic interpretation of some equations arising when the queueing system M/G/1 with egalitarian processor sharing is studied. To get this aim, we use, in addition to the new analytic method of the author, the well-known method of collective marks, in which two separate operations are involved: the “marking” of jobs, and the observation of “catastrophe” processes.     

An orchestration approach for unwanted Internet traffic identification

A simple examination of Internet traffic shows a wide mix of relevant and unwanted traffic. The latter is becoming increasingly harmful to network performance and service availability, while often consuming precious network and processing resources. Coordinated attacks, such as distributed denial-of-services (DDoS), large-scale scans, and worm outbreaks, occur in multiple networks simultaneously and become extremely difficult to detect using an individual detection engine. This paper presents the specification of a new orchestration-based approach to detect, and, as far as possible, to limit the actions of these coordinated attacks. Core to the proposal is a framework that coordinates the receiving of a multitude of alerts and events from detectors, evaluates this input to detect or prove the existence of anomalies, and consequently chooses the best action course. This framework is named Orchestration-oriented Anomaly Detection System (OADS). We also describe an OADS prototype implementation of the proposed infrastructure and analyze initial results obtained through experimentation with this prototype.