One big challenge in cloud-native network functions (CNFs) is the poor performance of packet forwarding. In this paper, we comprehensively analyze CPU cache usage in regard to the interprocess communication (vhost-user) as well as the packet I/O framework (DPDK). We cover four high-end CPUs, and examine 141 implementation/configuration patterns on an EIVU (Essential Implementation of Vhost-User) platform to clarify how the patterns affect the cache behaviors. The result pinpoints the roadblocks in a generalized form; cache invalidations stemming from three design/implementation factors are the major bottle-necks in vhost-user/DPDK, and shows potential of 100+ Mpps container networking for future software-centric environment.

Server-based NFV-nodes have disparate internals, such as simultaneous deployment of Virtual Network Functions (VNFs) and layered software abstractions including a virtual switch. The traditional operations tailored for function-hardware-coupled devices cannot cope with the increase of related components as well as complicated packet forwarding paths inside. Besides, self-development of VNFs attracting Telcos is still highly complicated work, due to lack of exact troubleshooting of internal NFV-nodes caused by exclusive resource management by Data-Plane Development Kit (DPDK). OPNFV Barometer provides means of stats acquisition, but internal figures of packet processing are still unveiled. In this paper, we propose an integrated metrics collection framework (NFV-VIPP) specialized to NFV-nodes. NFV-VIPP provides seamless understandings of system components in a node, and reveals the inside by transparently exposing implementation-related metrics. NFV-VIPP can be incorporated into Barometer/collected via RESTful APIs to reinforce system visibility, meaning that our framework bridges NFV-node internals to existing management frameworks. We explore NFV-node management using intra-VNF metrics obtained by NFVVIPP. Specifically, we prove that CPU-cycle consumption of inter-receive-polling is a driving force to estimate system load.

Recently, location service using Bluetooth Low-Energy (BLE) beacon is gaining popularity. There also exist researches that estimate the route of the user from the location estimation results, visualize, and analyze it. In the conventional route estimation method based on the BLE beacon, after estimating the all locations from the radio field strength of the BLE, the route is estimated from the sequence of the estimated locations. Therefore, one of the causes of deterioration in accuracy could be the fact that the estimation was repeated twice. Therefore, in this paper, we propose a novel estimation method of the user's route and stay using BLE beacons. Specifically, we apply the idea of global map matching, used in the field of GPS, to route estimation based on BLE beacons. Thus, the global optimum route in beacon network can be estimated from the radio field intensities of the BLE beacon directly. We associate a beacon network with path network to estimate the route and stay of users. We installed approximately 1,600 BLE beacon transmitters in all the classrooms and corridors of Nagoya Institute of Technology, and confirmed the effectiveness of the proposed method by experiments.

The Distributed Constraint Optimization Problem (DCOP) lies at the foundations of multiagent cooperation. With DCOPs, the optimization in distributed resource allocation problems is formalized using constraint optimization problems. The solvers for the problem are designed based on decentralized cooperative algorithms that are performed by multiple agents. In a conventional DCOP, a single objective is considered. The Multiple Objective Distributed Constraint Optimization Problem (MODCOP) is an extension of the DCOP framework, where agents cooperatively have to optimize simultaneously multiple objective functions. In the conventional MODCOPs, a few objectives are globally defined and agents cooperate to find the Pareto optimal solution. However, such models do not capture the interests of each agent. On the other hand, in several practical problems, the share of each agent is important. Such shares are modeled as preference values of agents. This class of problems can be defined using the MODCOP on the preferences of agents. In particular, we define optimization problems based on leximin ordering and Asymmetric DCOPs (Leximin AMODCOPs). The leximin defines an ordering among vectors of objective values. In addition, Asymmetric DCOPs capture the preferences of agents. Because the optimization based on the leximin ordering improves the equality among the satisfied preferences of the agents, this class of problems is important. We propose several solution methods for Leximin AMODCOPs generalizing traditional operators into the operators on sorted objective vectors and leximin. The solution methods applied to the Leximin AMODCOPs are based on pseudo trees. Also, the investigated search methods employ the concept of boundaries of the sorted vectors.

Distributed Constraint Optimization Problem (DCOP) has been studied as a fundamental component of multiagent systems. With DCOPs, various applications on multiagent systems are formalized as constraint optimization problems where variables and functions are distributed among agents. Leximin AMODCOP has been proposed as a class of Multiple Objective DCOPs, where multiple objectives for individual agents are optimized based on the leximin operator. This problem also relates to Asymmetric DCOPs based on its the criteria of fairness among agents. Previous studies explore only Leximin AMODCOPs on constraint graphs limited to functions with unary or binary scopes. We address the Leximin AMODCOPs on factor graphs that directly represent n-ary functions. A dynamic programming method on factor graphs is investigated as an exact solution method. In addition, for relatively dense problems, we also investigate several approximate/inexact algorithms.

A logical chaining of Virtual Network Functions (VNFs) enables toy-blocking style composition of various network services. VNFs are generally based on Virtual Machines (VMs) running on commodity servers. However, such VM-based VNFs have poor packet forwarding performance because of the virtual network I/O overhead, and therefore virtualization of the core network functions can degrade overall network performance. In this paper, we propose a novel packet aggregation method (PA-Flow) that aggregates packets having a same next-hop VNF with no further latency. Unlike existing packet aggregation approaches, our method is effective for the virtual network I/O to reduce packet processing cost inside the host. Moreover, aggregated packets are further aggregated as going through the service chain, which drastically reduces traffic amount at the core network. We have implemented a PA-Flow module that works between a virtual switch and QEMU under the DPDK/vhost-user framework. PA-Flow showed 170% higher throughput than that of the default DPDK/vhost-user based system for a 2-VNF service chain. In addition, latency was also improved by about 2 us because of the packet amount reduction.

The concept of network functions virtualization (NFV) has been embodied in commercial networks over the past years. Software-based virtual network functions have forwarding performance concerns in general, and various acceleration technologies have been developed so far, such as DPDK and vhost-user. Existence of several alternatives requires network engineers or operators to select appropriate technologies; however, no pragmatic criterion exists for constructing high-performance NFV-nodes. From their points of view, a lack of common benchmark and understanding of performance characteristics makes it difficult to predict hop-by-hop performance in a service chain, which results in prevention of NFV deployment in mission-critical networks. In this paper, we clarify performance characteristics of packet forwarding in NFV nodes focusing on three types of acceleration technologies; packet I/O architecture, virtual network I/O, and forwarding engine in a practical stage. We examined three packet I/O architectures (NAPI, netmap, and DPDK), three virtual I/O mechanisms (vhost-net, vhost-user, and SR-IOV), and four practical forwarding programs (Open vSwitch, OVS-DPDK, xDPd-DPDK, and Lagopus) with three referential programs (Linux Bridge, VALE, and L2FWD-DPDK). The experiment was conducted on a 40 GbE environment and we examined two device-under-test machines having different CPU performance. We argue performance characteristics of each technology and give quantitative analyses of the result. The key findings are: 1) CPU core speed has impact on both throughput and latency/jitter; 2) DPDK can allow performance prediction; 3) vhost-user is appropriate for real environment; and 4) OVS-DPDK provides a good combination of performance and functionality.

The notion of Software-Defined Networking (SDN) has already been introduced into cloud datacenter networks for provisioning virtual network environment. Network virtualization of today is generally achieved by L2-in-L3 tunneling protocols like VXLAN (Virtual eXtensible LAN) and NVGRE (Network Virtualization using Generic Routing Encapsulation) in public cloud datacenters. Some leading production packages for network virtualization have adopted an Edge-Overlay model that performs tunnel encapsulation and decapsulation processes at high-functional virtual switches to utilize existing network equipment. However, a severe performance problem arises because of the software-based tunneling processes. Alternatively, the STT (Stateless Transport Tunneling) protocol overcomes the problem by modifying the semantics of the TCP header, but such changes in semantics raises pragmatic issues in that network middleboxes can discard STT packets as an anomaly. In this paper, we propose a novel layer 4 protocol (Segment-oriented Connection-less Protocol, SCLP) for existing tunneling protocols such as VXLAN and NVGRE. SCLP is designed to not only accelerate the throughput of tunneling protocols, but prevent the packet discarding problem by providing a single-semantic header. Specifically, SCLP can exploit GRO (Generic Receive Offload) feature supported by the Linux kernel to reduce the number of packets to be software-interrupted. We implemented the SCLP protocol and applied it to the VXLAN protocol instead of UDP. As a result, the throughput of the VXLAN over SCLP protocol was almost doubled to the original UDP-based one at maximum.

Network Virtualization Overlays (NVO3) provides multi-tenancy services in cloud data centers with existing networking equipment. IP tunneling is an essential technology to logically separate each virtual traffic, in particular, Stateless Transport Tunneling (STT) is considered to achieve better performance using TCP Segmentation Offload (TSO) feature. Currently, there is no openly available implementation of STT, and its implementation and performance characteristics have not been studied in academic field so far. We have therefore implemented STT protocol and conducted performance evaluation by comparing with VXLAN protocol. In practice, the STT implementation has been done using a virtual NIC offloading framework, co-virtual switch (CVSW). CVSW is a software component that extends virtual NICs and provides high-level packet processing framework such as Open Flow Match-Action. In this paper, we describe implementation details of STT and performance evaluation results from various perspectives. The results showed that the actual performance of STT was almost equal to non-tunneling VM-to-VM communication and was two-times higher than that of VXLAN. Furthermore, we clarify the high-performance nature of STT is brought from both byte-stream characteristic of TCP and Generic Receive Offload (GRO) feature rather than widely believed TSO.

An Edge-Overlay model constructing virtual networks using both virtual switches and IP tunnels is promising in cloud datacenter networks. But software-implemented virtual switches can cause performance problems because the packet processing load is concentrated on a particular CPU core. Although multi queue functions like Receive Side Scaling (RSS) can distribute the load onto multiple CPU cores, there are still problems to be solved such as IRQ core collision of heavy traffic flows as well as competitive resource use between physical and virtual for packet processing. In this paper, we propose a software packet processing unit named VSE (Virtual Switch Extension) to address these problems by adaptively determining softirq cores based on both CPU load and VM-running information. Furthermore, the behavior of VSE can be managed by Open Flow controllers. Our performance evaluation results showed that throughput of our approach was higher than an existing RSSbased model as packet processing load increased. In addition, we show that our method prevented performance of high-loaded flows from being degraded by priority-based CPU core selection.

Distributed Constraint Optimization Problems (DCOPs) have been studied as fundamental problems in multiagent systems. The Max-Sum algorithm has been proposed as a solution method for DCOPs. The algorithm is based on factor graphs that consist of two types of nodes for variables and functions. While the Max-Sum is an exact method for acyclic factor-graphs, in the case that the factor graph contains cycles, it is an inexact method that may not converge. In this study, we propose a method that decomposes the cycles based on crossedged pseudo-trees on factor-graphs. We also present a basic scheme of distributed search algorithms that generalizes complete search algorithms on the constraint graphs and Max-Sum algorithm.

In current SDN paradigm, an edge-overlay (distributed tunneling) model using L2-in-L3 tunneling protocols, such as VXLAN, has attracted attentions for multi-tenant data center networks. The edge-overlay model can establish rapid-deployment of virtual networks onto existing traditional network facilities, ensure flexible IP/MAC address allocation to VMs, and extend the number of virtual networks regardless of the VLAN ID limitation. However, such model has performance and incompatibility problems on the traditional network environment. For L2 data center networks, this paper proposes a pure software approach that uses Open Flow virtual switches to realize yet another edge-overlay without IP tunneling. Our model leverages a header rewriting method as well as a host-based VLAN ID usage to ensure address space isolation and scalability of the number of virtual networks. In our model, any special hardware equipments like Open Flow hardware switch are not required and only software-based virtual switches and the controller are used. In this paper, we evaluate the performance of the proposed model comparing with the tunneling model using GRE or VXLAN protocol. Our model showed better performance and less CPU usage. In addition, qualitative evaluations of the model are also conducted from a broader perspective.

In this study, the authors forecast students’ future academic records using past attendance recording data and grade data. We use a Bayesian network as forecasting method. During construction of the Bayesian network forecasting model, unnecessary variables become noise and so lower the forecasting accuracy. Therefore, to improve the forecasting accuracy, we used information gain to reduce the number of variables in the model. As a result, accuracy improved.

We propose a method to solve problems that accompany recovering from operating system (OS) failures. First, to reduce recovery time, we make two OSes run simultaneously and configure them as an active-backup structure in one computer. This structure can provide a fast recovery from failures by a failover. Recovery time when using the proposed method is about 0.4 seconds at a minimum and up to about 10 seconds even if 2 GB memory is restored. Next, for smooth continuation of services after recovery, the proposed method preserves processes, their network connections, and file caches, and does not have runtime overhead to obtain a process execution status from the running active OS before a crash. In addition, the resources consumed to build the active-backup structure are only one CPU core and a small amount of memory. The hardware required to implement the proposed method is a multi-core processor and one disk for each OS, consequently, introduction of the proposed method incurs low cost. In the evaluation, we confirmed that the downtime was up to about 1.5 seconds when the active OS of the proposed system crashed while running a text editor, an NFS server, and a database server.

Anomaly-based Intrusion Prevention Systems have been studied to prevent zero-day attacks. However these existing systems can't prevent mimicry attacks because of the inadequacy of monitoring accuracy. Moreover, they provide no continuity for monitored applications when they have been compromised. In this paper, we propose a novel Intrusion Prevention System named Belem that detects anomaly states by checking the ordering of library functions and has a Continuing Execution Mechanism to provide application continuity. We implemented Belem on Linux and evaluated it.

We have proposed an auto-memoization processor. This processor automatically and dynamically memoizes both functions and loop iterations, and skips their execution by reusing their results. On the other hand, multi/many-core processors have come into wide use. The number of cores is expected to increase to a hundred or more. However, many programs do not have so much parallelism in them. Therefore it becomes very important to consider how to utilize many cores effectively. This paper describes a speedup technique for auto-memoization processor using speculative multi-threading. Two speculative threads will be forked on reuse test. The one assumes that the reuse test will succeed, and executes the following codes of the reuse target block speculatively. The other assumes that the reuse test will fail, and executes the reuse target block. These two threads conceal the overhead of auto-memoization processor. The result of the experiment with SPEC CPU95 suite benchmarks shows that proposing method improves the maximum speedup from 13.9% to 36.0%.

An immense amount of information is processed on the Internet due to its spread, increasing the severity of such problems as the disclosure of personal information; privacy protection is required. Research to protect anonymity has become crucial. Anonymous communication systems must consider a sudden breakaway of nodes. However existing systems isn't considering this enough. This paper proposes separating a node management layer from an anonymous communication layer. A novel anonymous communication system is realized by a node management layer that uses Chord, which is a distributed hash table, and the anonymous communication layer uses multiplex encryptions.

To speed up GA search, parallel distributed genetic algorithms are used. However in the current asynchronous parallel distributed genetic algorithm like the random-exchange or the sigma-exchange, it is hard to implement on the parallel computers or on the WS/PC clusters on the network, and it is easy to deadlock. We introduce an implementation method for asynchronous parallel distributed genetic algorithm by using the server-client model. In the proposed model, GA is executed on each client and each client communicates to only a server. Therefore, we need not to take care to synchronize between clients. This model is also safety and easy to implement. To evaluate proposed model, we applied to some problems, and confirm the effectiveness.

This paper introduces network based parallel genetic algorithm (NBPGA) we proposed before, and considers a reduction method of communication for NBPGA. NBPGA is implemented on ordinary single-processor computers, workstations or personal computers, by using client-server model. To evaluate NBPGA, we applied it to a few problems, and considered the reduction in communication quantity.

Parallel genetic algorithms are effective at solving large problems. Most are implemented on massively parallel computers and the efficiency depends on the parallel computing system. It is inappropriate to implement them on a distributed computing system connected by network. This paper proposes a client-server based approach to a parallel genetic algorithm with a delegate management model, that manages string exchange between subpopulations by the server and eliminates inter-subpopulation communications. It is easy to port and implement without any parallel computing systems. Some experiments solving the traveling salesman problem (100 cities) were carried out. The results show the effectiveness of the proposed model.

We propose an error correction technique for speaker-independent isolated word recognition by compensating for a word's likelihood. The likelihood is compensated for by the likelihood calculated by a phonetic bigram. The phonetic bigram is a phoneme model expressing the frame correlation within an utterance. A speaker-independent isolated word recognition experiment showed that our proposed technique reduces the recognition error compared to conventional techniques. The proposed technique achieves a performance almost equal to that without speaker adaptation compared to the conventional phoneme model adapted using several words.

A novel multiscale tree construction algorithm for three dimensional shape using energy controlled active balloon model (EC-ABM) is proposed. The key idea is that (1) internal and external energies and a number of surfaces which make up the ABM are controlled at each shrinking steps, and (2) each converged shape is regarded as one multiscale shape. Some experimental three dimensional recognition result of the proposed method with human face data are shown.

We proposed a distributed image processing environment VIOS. In this paper, the third version, VIOS III is proposed. In VIOS III, a new parallel processing language VPE-p which has flexible syntax for describing parallel algorithms has been developed. The new programmable buffer for accessing global variables through a local area network is also proposed. The description ability for parallel image processing algorithms and processing performance using workstation clusters and a multi processor system are investigated by several image processing and recognition algorithms.

Two kinds of compact and motor-less multimedia players are developed employing PC memory cards as the storage device. The VHS-quality video and the compact-disc quality sound play is supported. The MPEG-1 coding is incorporated for video and audio compression. Since the compressed information is stored in memory LSIs and moving mechanical components are not required, lightweight, shock proof and high-reliability are achieved. Instantaneous random program access is also available. A portable model, "Silicon View Player" equips with a 2.5-inch liquid crystal display panel and weights 295 g with a rechargeable battery. "Shopping Navigation" with a larger Pinch panel is a shopping guide terminal with an additional user-interactive function for sales promotion at retail stores. These models will pave the way for portable and business-use multimedia applications which do not use electro-magnetic storage devices.

We propose a method for recognition of a 3-D object using a multi scale description of the object and adaptive matching. MEGI model is a description model to represent arbitrary shapes. However, many MEGI elements are necessary to represent uneven or carved surfaces with accuracy, so it is difficult to use them for recognition. As a solution, we make a tree which corresponds to the multi scale description of the object. While tracing the tree from the root which corresponds the coarsest representation to leaves, a matching algorithm presented in this paper assigns a different scale to each part of the object adaptively and estimates the matching score effectively.

Description and recognition of objects is the central considerations of research on computer vision. The key issue is how to represent 3D objects on a machine for recognizing them. Researchers of computer vision commonly employ the extended Gaussian image (EGI) model, but it is not able to express concave objects. In this paper, an MEGI(more EGI) model and coefficient of extended spherical correlation have been proposed. The MEGI model is an, extended EGI modeling which is able to represent concave objects. The extended spherical correlation is a measure for recognizing objects using the MEGI model. It has been demonstrated that this model is able to recognize 3D objects, including concave ones, and to distinguish objects using a part of MEGI from range data.

Conventional X-ray tomosynthesis with film can provide a sagittal slice image with a single scan. This technique has the advantage of enabling reconstruction of a sagittal slice which is difficult to obtain from the X-ray CT system. However, only an image on the focal plane is obtained by a single scan. Furthermore, the image is degraded by superimpositions of the structures outside of the focal plane. A new three-dimensional image reconstruction method is proposed. This method utilizes a three-dimensional convolution process with an inverse filter function which is derived analytically by the point spread function of the projection and backprojection geometry. A digital tomosynthesis system has also been constructed for the purpose of evaluating the proposed method. This system was used in phantom experiments and clinical evaluations, and it was confirmed that the proposed method was able to reconstruct a better three-dimensional image with less artifacts from outside of the focused slice.