Nobuharu Kami

Multi-root Share of Single-Root I/O Virtualization (SR-IOV) Compliant PCI Express Device

We have achieved sharing a single-root I/O virtualization (SR-IOV) compliant PCI Express (PCIe) I/O device among multiple computers. A device share not only inside a single computer among virtual machines, but also among multiple computers attracts a great interest because it provides efficient utilization of computer resources. Because PCIe is originally a single-root system, realizing multi-root I/O virtualization is much more difficult than SR-IOV. We allocate virtual instances called VF of an SR-IOV-compliant I/O device to an individual computer by virtualizing IOV configuration and translating memory address of VF to that of the allocated computer. With the FPGA implementation, we have achieved sharing a commercially-available network interface card among three computers without modification in OS/driver and device itself. In addition, the performance reaches 99% of the device in the best case whereas the implementation is in the early stage, indicating this method provides not only MR sharing but high performance at the same time.

Optical interconnection as an IP macro of a CMOS library

With the aim of producing a switch LSI with several-hundred-Gb/s-class throughput, we first developed LAN-LSIs with 10-Gb/s-class optical input/output (I/O) buffer. The optical I/O buffer was supplied as a set consisting of an optical-electrical package, soft or hard macros of the optical interface, and library files for a targeted CMOS technology. Using such library files, system LSI designers can implement an optical interface in their chips in a conventional design manner.

An analog front-end chip set employing an electro-optical mixed design on SPICE for 5-Gb/s/ch parallel optical interconnection

A chip set composed of a laser-diode driver (LDD) and an optical receiver (RCV), which incorporates a full 2D (reshape, regenerate) function, has been developed by using silicon bipolar technology for a four-channel 5-Gb/s parallel optical transceiver. An electro-optical mixed design on SPICE of the LDD and the LD is accomplished by describing the rate equations of the LD as an electrical circuit. This design accommodates easy connectivity of the LDD chip to the LD in the optical transmitter module without the need for adjustment of the optical waveform. A pseudobalanced transimpedance amplifier (TIA) and feedforward automatic decision threshold control (ATC) in the RCV minimize the number of off-chip bypass capacitors, eliminate the need for any off-chip coupling capacitors, and keep crosstalk less than -50 dB and low cutoff frequency less than 80 kHz. A prototype parallel optical transmitter module and a prototype receiver module, based on the chip set, demonstrated asynchronous four-channel 5-Gb/s operation. The chip set has a throughput of 20 Gb/s with a power dissipation of 1.3 W at a 3.3-V supply.

Algorithm for detecting significant locations from raw GPS data

We present a fast algorithm for probabilistically extracting significant locations from raw GPS data based on data point density. Extracting significant locations from raw GPS data is the first essential step of algorithms designed for location-aware applications. Assuming that a location is significant if users spend a certain time around that area, most current algorithms compare spatial/temporal variables, such as stay duration and a roaming diameter, with given fixed thresholds to extract significant locations. However, the appropriate threshold values are not clearly known in priori and algorithms with fixed thresholds are inherently error-prone, especially under high noise levels. Moreover, for N data points, they are generally O(N2) algorithms since distance computation is required. We developed a fast algorithm for selective data point sampling around significant locations based on density information by constructing random histograms using locality sensitive hashing. Evaluations show competitive performance in detecting significant locations even under high noise levels.

Smart M2M Data Filtering Using Domain-Specific Thresholds in Domain-Agnostic Platforms

Due to the demand for homogeneous, intelligent, and automated access to data measured anywhere and from any device, Machine-to-Machine (M2M) platforms are evolving as globally-intended multi-layer solutions that provide such access, abstracting from all technology-specific tasks. In order to preserve the stability of their potentially huge data-handling systems and the usefulness of their Big Data, M2M platforms must maintain some data selection and filtering logic. A challenge that appears in modern M2M platforms is related to the decoupling of the front end (devices, area networks) from the backend (applications, databases). Because of this decoupling, domain-specific tricks cannot be applied any more for filtering at the front end. This paper presents a solution using domain-specific filtering thresholds in a domain-agnostic platform, as well as filtering flows and algorithms tailored to modern M2M platforms. Their combination assembles the first filtering solution that supports the unified handling of heterogeneous filters. In an evaluation from the utility-monitoring domain, instances of our approach showed high efficiency of configuration and were the only ones to achieve, for example, forwarding less than 25% of the captured data maintaining a coverage ratio bigger than 50% for all considered applications.

A tracking area list configuration method to mitigate burst location updates

The quality of the services provided by cellular networks is negatively affected by the mobility management signal burst traffic caused by simultaneous movement of user equipment in mass transportation vehicles. Long Term Evolution networks track the location of a user equipment at the granularity of a tracking area list (TAL). This strategy supplies a flexible mechanism for reducing and distributing the signaling loads for mobility management. However, it is not clear how to effectively mitigate the burst location update signals caused by mass transportation systems by taking advantage of TALs. Therefore, we propose a tracking area list configuration method by solving an optimization problem. The proposed method detects the cells with burst location updates from the location update records and generates rules for the TAL configuration that selectively mitigates the burst location updates.

Scalable and reliable platform for service-oriented networking and computing systems

We propose a reliable, scalable platform that employs a reliability management framework to support service-oriented networking and computing systems. Although virtualization is widely used to make systems reconfigurable, it fails to provide a totally integrated management framework in which the allocation of required services to adequate physical resources takes into account the presence of shared risk groups. We present basic system architecture and a highly scalable hardware-platform, as well as a reliability management framework that provides both a service availability management interface and automated risk assessment. Our evaluation of a prototype system shows that with the proposed framework, complete service breakdown will occur only in the case of multiple-point failure. While this corresponds to the same level of reliability as ordinary manual risk management, resource utilization efficiency is twice as high

QoI-Based Data Upload Control for Mobility-Aware Cloud Services

Mobility-aware cloud services such as fleet management systems need to understand the positions of mobile devices accurately in a real-time manner. Generally speaking, positioning accuracy and data traffic load are in a trade-off relation. Highly accurate real-time positioning requires frequent location data upload and hence results in heavy data traffic load. Although not all data are equally important, data of low importance often consumes a lot of network resources. This paper presents a data upload control method that the dynamically assesses quality of information (QoI) of measured data at mobile devices. The proposed method balances high accuracy with low traffic loads to achieve efficient vehicle position management. We evaluated the performance of the proposed method using both artificial and actual GPS data and confirmed that it successfully controlled the accuracy and network traffic load according to application requirements.

Efficient filtering processes for machine-to-machine data based on automation modules and data-agnostic algorithms

Machine-to-machine (M2M) platforms are evolving as large-scale multi-layer solutions that unify the access and the control of all devices that are being equipped with the capability to perform automated tasks and to report data based on connectivity to a backend system. As the integration of more and more devices in such platforms results in the need to handle big M2M data, M2M platforms need to automate their configuration and include appropriate data filtering frameworks and algorithms. Otherwise, the collected raw data can become expensive, unmanageable, and of low quality. This paper presents how data filtering processes can be automated as part of an M2M self-configuration framework and describes a solution that enables the seamless adjustment of domain-specific filtering thresholds in domain-agnostic platforms, based on quality-of-information calculations and M2M-specific data categorisation. An evaluation from the facilities-monitoring domain shows that our approach was the only one to achieve, for example, forwarding less than 25% of the monitored data maintaining at the same time a coverage ratio bigger than 50% for all considered applications. Further, a projection of this evaluation to a Smart City scale indicates that such gains can make database queries up to many seconds faster.

Adaptive mobility management in cellular networks with multiple model-based prediction

With the spread of smartphones and machine-to-machine devices that communicate via cellular networks, mobility and communication characteristics in cellular networks have become diverse. However, because existing cellular networks are designed for human-to-human communication, mobility management in cellular networks is not necessarily effective. In this paper, we propose an adaptive method of designing registration and paging areas to reduce costs in the mobility management of cellular networks. The proposed method uses the interacting multiple model algorithm to support multiple motion models that switch over time.