Toyofumi Takenaka

Adaptive load balancing content address hashing routing for reverse proxy servers

We propose a novel routing algorithm for reverse proxy servers, called adaptive loud balancing content address hashing (AH), and evaluate the performance of the proposed routing algorithm compared with that of the content address hashing (SH) and the hash and slide (HS) routing algorithms. The proposed AH routing algorithm calculates the popularity of pages in the load balancer using an LFU caching technique and periodically makes a popularity list. Using this popularity list, the proposed routing algorithm selects a reverse proxy server as follows. When the requested page appears in the popularity list, the request is routed according to the round robin method; otherwise, it is routed according to the content address hashing method. We evaluate and compare the AH, SH and HS routing algorithms by simulation experiments from the viewpoints of load balancing, consumed cache space and cache hit rate. Simulation experiments show that the proposed AH routing algorithm achieves almost the same degree of load balancing as the HS algorithm and the same cache hit rate as the SH algorithm, for reverse proxy servers in various web site environments.

Optimization of input admissible traffic flow guaranteeing QoS

We propose an admission control method at edge nodes. This method aims at maximizing network throughput and calculates an input traffic limitation ratio at each edge node, which is used for admission control. Since the proposed method only uses input traffic volume information, it does not need to take into account the traffic matrix among edge nodes. Therefore, this method can be easily applied to actual networks in a practical way. We formulate a linear programming (LP) to solve the optimization problem of the input traffic limitation ratio of each node for any traffic matrix. This problem cannot be solved as it is, so we modify the LP problem by applying a duality theorem in the field of LP problems.

PARC: Power aware routing and clustering scheme for wireless sensor networks

In wireless sensor networks constructed from battery driven nodes, it is difficult to supply electric power to the nodes. Because of this, the power consumption must be reduced. To cope with this problem, clustering techniques have been proposed. EACLE is a method that uses a clustering technique. In EACLE, the route selection is executed independently after the CH (Cluster Head) selection. These two phase controls increase overheads and reduce the battery power, which leads to shorten the lifetime of wireless sensor networks. This paper proposes a novel routing and clustering method called PARC for wireless sensor networks that reduces these overheads by integrating the cluster selection phase and the route construction phase into a single phase. However, PARC has a following weak point that the batteries of CHs around the sink node are depleted earlier than the other nodes and the sink node cannot collect sensing data. This phenomenon is called a hot spot problem. In order to cope with this problem of PARC, we propose PARC+, which extends the CH selection method of PARC such as more nodes around the sink can be selected as a CH node. We evaluate our proposed methods by simulation experiments and show its effectiveness.

Improvement of burst transmission delay using offset time for burst assembly in optical burt switching

We propose a novel algorithm for reducing burst transmission delay and packet loss rate. In optical burst switching (OBS), a certain offset time and burst assembly period are needed, leading to transmission delay. In order to reduce the delay, the proposed method transmits a control packet before the burst assembly period expires. Due to this mechanism, a method is needed to estimate burst length. To estimate burst length, the proposed method adopts the Jacobson/Karels algorithm used in TCP. Moreover, to reduce the estimation error, the proposed method uses a head-dropping algorithm for the burst collision procedure. We evaluate packet delay, packet loss rate and estimation error rate by simulation experiments. The simulation results show that the proposed method has a lower packet loss rate and delay than the conventional method and burst estimation error can be neglected.

WDM Multicast Tree Construction Algorithms for Minimizing Blocking Probability under a Delay Constraint

We propose novel tree construction algorithms for multicast communication in photonic networks. Since multicast communications consume many more link resources than unicast communications, effective algorithms for route selection and wavelength allocation are required. We propose novel tree construction algorithms called the weighted Steiner tree (WST) algorithm and the composite weighted Steiner tree (CWST) algorithm. Since the proposed Weighted Steiner Tree algorithm takes account of the number of wavelengths available for each link, this algorithm can select those links which have larger number of available wavelengths. By using this link selection technique, this algorithm can accept many more multicast requests than other multicast tree construction algorithms. However, under certain delay constraints, the blocking characteristics of the proposed weighted Steiner tree algorithm deteriorate since some light paths between source and destination use long hops and cannot satisfy the delay constraint. In order to adapt the approach to the delay-sensitive environments, we have devised the composite weighted Steiner tree algorithm comprising the weighted Steiner tree algorithm and the Dijkstra algorithm for use in a delay constrained environment such as an IPTV application. In this paper, we also give the results of simulation experiments which demonstrate the superiority of the proposed composite weighted Steiner tree algorithm compared with the distributed minimum hop tree (DMHT) algorithm, from the viewpoint of the light-tree request blocking probability.

Overlap Degree Aware Routing in All-Optical Routing Networks

Dynamic routing and wavelength assignment (RWA) is an attractive method for the efficient use of network resources in all-optical networks. We present a novel fixed alternate routing method referred to as Overlap-Degree Aware (ODA) routing in all-optical networks. A lot of researchers have focused on the shortest path routing and alternate shortest path routing taking into acount link and wavelength usage so as to reduce the consumption of network resources. The authors, however, believe that in order to minimize the blocking probability, it is important to consider not only the consumption of link and wavelength resources but also the existence of the other flows when a routing decision is made. The ODA routing decides routes using the knowledge of ingress-egress node pairs, and tries to prevent future path requests from being blocked unnecessarily by reserving link and wavelength resources for the future requests. Our simulation results show that our new routing algorithm outperforms Fixed-Alternate Routing (FAR) and Weighted Least Congestion Routing (WLCR) from the viewpoint of call blocking probability.

High-speed calculation method for large-scale multi-layer network design problem

Multi-layer network optimization has been studied for efficient use of network resources by solving Mixed Integer Linear Programming (MILP) problem. Here, the multi-layer network consists of lambda-layer network and IP-layer network. However, when applying this MILP problem for a large-scale network with a few hundred nodes, huge amount of variables are used and cannot be handled by middle size computers. In order to cope with this problem, some heuristic methods have been proposed and evaluated from viewpoints of accuracy and calculation time for their computation results. These heuristic methods reduce variables at lambda-layer and also decompose the original MILP problem into smaller sub-problems and induce an approximate solution by solving these smaller MILP problems. However, these heuristic methods cannot obtain accurate results in many cases. In this paper, we propose a novel variable reduction method at both lambda-layer and IP-layer by excluding long hop routes. We evaluate our proposed method and the conventional methods and show the availability of our method.

High-speed topology construction method for network power saving

Recently a drastic increase of power consumption of network devices has been big problem in the green ICT field. In order to cope with this problem, some power saving method has been proposed, where traffic aggregations are executed and the powers of unnecessary links are turned off. In this method, every links are tried to be turned off and are checked whether traffic congestions are caused or not and whether network disconnections occur or not. However, this method causes frequent change of link on/off and induces network instabilities in the actual network environment, where traffic conditions change in real time. Moreover, in this method, since the shortest hop routing has been adopted, traffic congestions might occur in specific links. In this paper, in order to cope with these problems, we propose a High-Speed Topology Construction method, where an initial topology is constructed by a minimum spanning tree algorithm in order to maximize sleep links. When traffic congestions are caused, a proper link addition proposed in this paper is executed in order to resolve traffic congestions. We evaluate our proposed methods by simulation experiments and show its effectiveness.

OPN09-01: Overlap-Degree Aware Routing in All-Optical Networks

Dynamic routing and wavelength assignment (RWA) is an attractive method for the efficient use of network resources in all-optical networks. We present a novel fixed alternate routing method referred to as overlap-degree aware (ODA) routing in all-optical network. The authors believe that in order to minimize the blocking probability, it is important to consider not only the consumption of wavelength resources but also the existence of the other flows when a routing decision is made. Our simulation results show that our new routing algorithm outperforms fixed-alternate routing (FAR) and weighted least congestion routing (WLCR) from the viewpoint of call blocking probability.

Load Balancing Routing Algorithm for Reverse Proxy Servers

We propose a novel routing algorithm for reverse proxy servers, called load balancing content address hashing (LB-CAH), and evaluate the performance of the proposed routing algorithm compared with that of the content address hashing (CAH) and the hash and slide (HAS) routing algorithms. The proposed LB-CAH routing algorithm calculates the popularity of pages in the load balancer using an LFU caching technique and periodically makes a popularity list. Using this popularity list, the proposed routing algorithm selects a reverse proxy server as follows. When the requested page appears in the popularity list, the request is routed according to the round robin method; otherwise, it is routed according to the content address hashing method. We evaluate and compare the LB-CAH, CAH and HAS routing algorithms by simulation experiments from the viewpoints of load balancing, consumed cache space and cache hit rate. Simulation experiments show that the proposed LB-CAH routing algorithm achieves almost the same degree of load balancing as the HAS algorithm and the same cache hit rate as the CAH algorithm for reverse proxy servers in various web site environments.