Chase Q. Wu

Experimental Analysis of File Transfer Rates over Wide-Area Dedicated Connections

File transfers over dedicated connections, supported by large parallel file systems, have become increasingly important in high-performance computing and big data workflows. It remains a challenge to achieve peak rates for such transfers due to the complexities of file I/O, host, and network transport subsystems, and equally importantly, their interactions. We present extensive measurements of disk-to-disk file transfers using Lustre and XFS file systems mounted on multi-core servers over a suite of 10 Gbps emulated connections with 0–366 ms round trip times. Our results indicate that large buffer sizes and many parallel flows do not always guarantee high transfer rates. Furthermore, large variations in the measured rates necessitate repeated measurements to ensure confidence in inferences based on them. We propose a new method to efficiently identify the optimal joint file I/O and network transport parameters using a small number of measurements. We show that for XFS and Lustre with direct I/O, this method identifies configurations achieving 97% of the peak transfer rate while probing only 12% of the parameter space.

Profiling transport performance for big data transfer over dedicated channels

The transfer of big data is increasingly supported by dedicated channels in high-performance networks. Transport protocols play a critical role in maximizing the link utilization of such high-speed connections. We propose a Transport Profile Generator (TPG) to characterize and enhance the end-to-end throughput performance of transport protocols. TPG automates the tuning of various transport-related parameters including socket options and protocol-specific configurations, and supports multiple data streams and multiple NIC-to-NIC connections. To instantiate the design of TPG, we use UDT as an example in the implementation and conduct extensive experiments of big data transfer over high-speed network channels to illustrate how existing transport protocols benefit from TPG in optimizing their performance.

Measurement-based performance profiles and dynamics of UDT over dedicated connections

Wide-area data transfers in high-performance computing and big data scenarios are increasingly being carried over dedicated network connections that provide high capacities at low loss rates. UDP-based transport protocols are expected to be particularly well-suited for such transfers but their performance is relatively unexplored over a wide range of connection lengths, compared to TCP over shared connections. We present extensive throughput measurements of UDP-based Data Transfer (UDT) over a suite of physical and emulated 10 Gbps connections. In sharp contrast to current UDT analytical models, these measurements indicate much more complex throughput dynamics that are sensitive to the connection modality, protocol parameters, and round-trip times. Lyapunov exponents estimated from the Poincaré maps of UDT traces clearly indicate regions of instability and complex dynamics. We propose a simple model based on the ramp-up and sustainment regimes of a generic transport protocol, which qualitatively illustrates the dominant monotonicity and concavity properties of throughput profiles and relates them to Lyapunov exponents. These measurements and analytical results together enable us to comprehensively evaluate UDT performance and select parameters to achieve high throughput, and they also provide guidelines for designing effective transport protocols for dedicated connections.

Profiling Optimization for Big Data Transfer over Dedicated Channels

The transfer of big data is increasingly supported by dedicated channels in high-performance networks, where transport protocols play an important role in maximizing application-level throughput and link utilization. The performance of transport protocols largely depend on their control parameter settings, but it is prohibitively time consuming to conduct an exhaustive search in a large parameter space to find the best set of parameter values. We propose FastProf, a stochastic approximation-based transport profiler, to quickly determine the optimal operational zone of a given data transfer protocol/method over dedicated channels. We implement and test the proposed method using both emulations based on real-life performance measurements and experiments over physical connections with short (2ms) and long (380ms) delays. Both the emulation and experimental results show that FastProf significantly reduces the profiling overhead while achieving a comparable level of end-to-end throughput performance with the exhaustive search-based approach.

Improved multi-resolution method for MLE-based localization of radiation sources

Multi-resolution grid computation is a technique used to speed up source localization with a Maximum Likelihood Estimation (MLE) algorithm. In the case where the source is located midway between grid points, the MLE algorithm may choose an incorrect location, causing following iterations of the search to close in on an area that does not contain the source. To address this issue, we propose a modification to multi-resolution MLE that expands the search area by a small percentage between two consecutive MLE iterations. At the cost of slightly more computation, this modification allows consecutive iterations to accurately locate the target over a larger portion of the field than a standard multi-resolution localization. The localization and computation performance of our approach is compared to both standard multi-resolution and single-resolution MLE algorithms. Tests are performed using seven data sets representing different scenarios of a single radiation source located within an indoor field of detectors. Results show that our method (i) significantly improves the localization accuracy in cases that caused initial grid selection errors in traditional MLE algorithms, (ii) does not have a negative impact on the localization accuracy in other cases, and (iii) requires a negligible increase in computation time relative to the increase in localization accuracy.

Experiments and Analyses of Data Transfers over Wide-Area Dedicated Connections

Dedicated wide-area network connections are increasingly employed in high-performance computing and big data scenarios. One might expect the performance and dynamics of data transfers over such connections to be easy to analyze due to the lack of competing traffic. However, non-linear transport dynamics and end-system complexities (e.g., multi-core hosts and distributed filesystems) can in fact make analysis surprisingly challenging. We present extensive measurements of memory-to-memory and disk-to-disk file transfers over 10~Gbps physical and emulated connections with 0-366~ms round trip times (RTTs). For memory-to-memory transfers, profiles of both TCP and UDT throughput as a function of RTT show concave and convex regions; large buffer sizes and more parallel flows lead to wider concave regions, which are highly desirable. TCP and UDT both also display complex throughput dynamics, as indicated by their Poincare maps and Lyapunov exponents. For disk-to-disk transfers, we determine that high throughput can be achieved via a combination of parallel I/O threads, parallel network threads, and direct I/O mode. Our measurements also show that Lustre filesystems can be mounted over long-haul connections using LNet routers, although challenges remain in jointly optimizing file I/O and transport method parameters to achieve peak throughput.

Network detection of radiation sources using ROSD localization

Networks of radiation counters are increasingly being deployed in monitoring applications to provide faster and better detection than individual detectors. Their performances critically depend on the algorithms used to aggregate measurements from individual detectors. Recently, localization-based algorithms have been developed for network detection, where multiple source location estimates are generated based on the measurements from various “dispersed” subnets: i) when a source is present, these source location estimates form a single dominant cluster; ii) otherwise, they are spatially dispersed. For example, the triangulation-based detection method [1] employs a closed-form quadratic expression for source location estimates using a subnet of three detectors. This method works well in relatively simple detector configurations, but may exhibit unpredictable performances in complex settings mainly due to the increased number of imaginary roots in the closed-form solution.

Data Transfer Advisor with Transport Profiling Optimization

The network infrastructures have been rapidly upgraded in many high-performance networks (HPNs). However, such infrastructure investment has not led to corresponding performance improvement in big data transfer, especially at the application layer, largely due to the complexity of optimizing transport control on end hosts. We design and implement ProbData, a PRofiling Optimization Based DAta Transfer Advisor, to help users determine the most effective data transfer method with the most appropriate control parameter values to achieve the best data transfer performance. ProbData employs a profiling optimization-based approach to exploit the optimal operational zone of various data transfer methods in support of big data transfer in extreme-scale scientific applications. We present a theoretical framework of the optimized profiling approach employed in ProbData as well as its detailed design and implementation. The advising procedure and performance benefits of ProbData are illustrated and evaluated by proof-of-concept experiments in real-life networks.

On Routing of Multiple Concurrent User Requests in Multi-Radio Multi-Channel Wireless Mesh Networks

Multiple-radio multiple-channel (MRMC) wireless mesh networks (WMNs) generally serve as wireless backbones for ubiquitous Internet access. These networks often face a challenge to satisfy multiple user traffic requests simultaneously between different source-destination pairs with different data transfer requirements. We construct analytical network models and formulate such multi-pair routing as a rigorous optimization problem. We design a cooperative routing and scheduling algorithm with channel assignment, in which, a primary path is built upon the selection of appropriate link patterns. The performance of the proposed algorithm is illustrated by simulation-based combinatorial experiments.

A source-attractor approach to network detection of radiation sources

Radiation source detection using a network of detectors is an active field of research for homeland security and defense applications. We propose Source-attractor Radiation Detection (SRD) method to aggregate measurements from a network of detectors for radiation source detection. SRD method models a potential radiation source as a “magnet”-like attractor that pulls in pre-computed virtual points from the detector locations. A detection decision is made if a sufficient level of attraction, quantified by the increase in the clustering of the shifted virtual points, is observed. Compared with traditional methods, SRD has the following advantages: i) it does not require an accurate estimate of the source location from limited and noise-corrupted sensor readings, unlike the localization-based methods, and ii) its virtual point shifting and clustering calculation involve simple arithmetic operations based on the number of detectors, avoiding the high computational complexity of grid-based likelihood estimation methods. We evaluate its detection performance using canonical datasets from Domestic Nuclear Detection Offices (DNDO) Intelligence Radiation Sensors Systems (IRSS) tests. SRD achieves both lower false alarm rate and false negative rate compared to three existing algorithms for network source detection.