This letter proposes a hierarchical label-switched path (LSP) setup scheme, called ConSet, for multi-layer generalized multi-protocol label switching (GMPLS) networks. ConSet allows a Path message to be transmitted to the downstream neighbor node without waiting for the establishment of the higher-order LSP. Confirmation of the establishment of the higher-order LSP is performed at the ingress node of the higher-order LSP before a Resv message of the lower-order LSP is transmitted to the upstream neighbor node. ConSet is able to set up hierarchical LSPs faster than the sequential scheme.

This paper proposes an algorithm for calculating routes that considers the include route constraint while minimizing cost. A route with include route constraint has to traverse a group of assigned nodes. The trouble when calculating a route that satisfies an include route constraint is that routes set in different sections may traverse the same link. In order to prevent this violation (overlap), we introduce an alternate route selection policy. Numerical results show that the probability of finding appropriate routes (no overlap) is more than 95% with the proposed algorithm while only 35% with the conventional algorithm.

We propose a mechanism called “optical plug and play” for constructing GMPLS networks automatically. It offers lower operation effort and fast network construction, and avoids misconfiguration. Optical plug and play architecture has its procedure, a link-up search mechanism for OXCs, network and node architectures to realize optical plug and play, and an LMP extension to exchange the information between nodes necessary for identifying adjacent nodes. We implement prototypes of both OXCs and routers that support the optical plug and play proposal. Simulations and experiments confirm its performance and feasibility.

In recent years, the time variation of Internet traffic has increased due to the growth of streaming and cloud services. Backbone networks must accommodate such traffic without congestion. Traffic engineering with traffic prediction is one approach to stably accommodating time-varying traffic. In this approach, routes are calculated from predicted traffic to avoid congestion, but predictions may include errors that cause congestion. We propose prediction-based traffic engineering that is robust against prediction errors. To achieve robust control, our method uses model predictive control, a process control method based on prediction of system dynamics. Routes are calculated so that future congestion is avoided without sudden route changes. We apply calculated routes for the next time slot, and observe traffic. Using the newly observed traffic, we again predict traffic and re-calculate the routes. Repeating these steps mitigates the impact of prediction errors, because traffic predictions are corrected in each time slot. Through simulations using backbone network traffic traces, we demonstrate that our method can avoid the congestion that the other methods cannot.

This proposes a scalable QoS control scheme, called Elephant Flow Control Scheme (EFCS) for high-speed large-capacity networks; it controls congestion and provides appropriate bandwidth to normal users' flows by controlling just the elephant flows. EFCS introduces a sampling packet threshold and drops packets considering flow size. EFCS also adopts a compensation parameter to control elephant flows to an appropriate level. Numerical results show that the sampling threshold increases control accuracy by 20% while reducing the amount of memory needed for packet sampling by 60% amount of memory by packet sampling; the elephant flows are controlled as intended by the compensation parameter. As a result, EFCS provides sufficient bandwidth to normal TCP flows in a scalable manner.

This paper proposes the Path Computation Element (PCE)-based backbone network architecture and verifies its feasibility through implementation and experiments. PCE communication Protocol (PCEP) is implemented for communication between the PCE and the management system to control and manage Generalized Multi-Protocol Label Switching (GMPLS)-based backbone networks.

We propose ATM switching nodes with a feedback rate control scheme, AREX, which does not require a large buffer space and does not deteriorate throughput even in large-scale and high-speed ATM-WANs. The goal of our study is to establish the ATM multi-protocol emulation network ALPEN, which is an ATM-WAN architecture for establishing a backbone for multimedia networks. ALPEN achieves an ATM-WAN which is robust against long propagation delays. It also provides high performance without a large buffer space in an ATM-WAN environment. In ALPEN, each transit node informs the edge nodes only its residual bandwidth ratio. The edge nodes support multiple ATM-layer services by emulating them based on the information notified by transit nodes. Our research has been directed towards achieving high performance ABR (Available Bit Rate) service in an ATM-WAN by using ALPEN. The conventional ABR service requires transit nodes to have relatively high calculation power and large buffer space to overcome the effect of the long propagation delays common in WANs. ALPEN node systems have been developed for trials with actual network traffic. ALPEN with AREX reduces the calculation load of transit nodes for ABR service. That is confirmed by the size of the DSP program created for a test system. ALPEN with AREX is, therefore, able to emulate ABR service with higher performance in ATM-WANs, because ALPEN edge nodes are able to indicate the users allowed by ER (Explicit Rate) feedback. The network throughput, maximum queue length at congestion point, and burst transmission rate are determined by simulation. ALPEN with AREX achieves better performances than the conventional ABR network.

IP Fast Reroute techniques have been proposed to achieve fast failure recovery, just a few milliseconds. The basic idea of IP Fast Reroute is to reduce recovery time by precomputing backup routes. The multiple routing configurations (MRC) algorithm was proposed to implement IP Fast Reroute. MRC prepares backup configurations, which are used for finding a detour route after a failure. However, this algorithm establishes too many backup configurations to recover from failures. We propose a new backup configuration computation algorithm that creates the fewest possible configurations. The basic idea is to construct a spanning tree that excludes failure links in each backup configuration. We show that the effectiveness of our algorithm is especially high in large-scale power-law networks.

Generalized Multi-Protocol Label Switching (GMPLS) is being developed in the Internet Engineering Task Force (IETF). In GMPLS-based wavelength-division-multiplexing (WDM) optical networks, a wavelength in a fiber is used as a label. In the existing GMPLS signaling protocol for bidirectional paths in WDM networks with the wavelength continuity constraint, bidirectional path setup fails with high probability because the upstream label allocated by the previous hop node may not be accepted at the transit node. To solve this problem, this paper proposes an efficient bidirectional label switched path (LSP) setup scheme based on an upstream label set. Called the Upstream Label Set (ULS) scheme, it is an extension of the existing GMPLS signaling protocol. The ULS scheme is consistent with the existing GMPLS signaling procedure and so offers backward compatibility. The numerical results suggest that when the number of the LSP setup retries is limited, the ULS scheme offers lower blocking probability than the existing GMPLS signaling scheme which uses only with the upstream label (UL). In addition, under the condition that the constraint of the number of LSP setup retries is relaxed, the LSP setup time of the ULS scheme is faster than that of the existing scheme. Furthermore, by using our developed prototype of the GMPLS control system, in which the ULS scheme was installed, we demonstrated that the ULS scheme successfully setup bidirectional LSPs.

We survey traffic matrix models, whose elements represent the traffic demand between source-destination pair nodes. Modeling the traffic matrix is useful for multilayer Traffic Engineering (TE) in IP optical networks. Multilayer TE techniques make the network so designed flexible and reliable. This is because it allows reconfiguration of the virtual network topology (VNT), which consists of a set of several lower-layer (optical) paths and is provided to the higher layer, in response to fluctuations (diurnal) in traffic demand. It is, therefore, important to synthetically generate traffic matrices as close to the real ones as possible to maximize the performance of multilayer TE. We compare several models and clarify their applicability to VNT design and control. We find that it is difficult in practice to make an accurate traffic matrix with conventional schemes because of the high cost for data measurement and the complicated calculations involved. To overcome these problems, we newly introduce a simplified traffic matrix model that is practical; it well mirrors real networks. Next, this paper presents our developed server, the IP Optical TE server. It performs multilayer TE in IP optical networks. We evaluate the effectiveness of multilayer TE using our developed IP Optical server and the simplified traffic matrix. We confirm that multilayer TE offers significant CAPEX savings. Similarly, we demonstrate basic traffic control in IP optical networks, and confirm the dynamic control of the network and the feasibility of the IP Optical TE server.

A call admission control (CAC) scheme based on statistical information is proposed, called the statistical CAC scheme. A conventional scheme needs to manage session information for each link to update the residual bandwidth of a network in real time. This scheme has a scalability problem in terms of network size. The statistical CAC rejects session setup requests in accordance to a pre-computed ratio, called the rejection ratio. The rejection ratio is computed by using statistical information about the bandwidth requested for each link so that the congestion probability is less than an upper bound specified by a network operator. The statistical CAC is more scalable in terms of network size than the conventional scheme because it does not need to keep accommodated session state information. Numerical results show that the statistical CAC, even without exact session state information, only slightly degrades network utilization compared with the conventional scheme.

This paper evaluates three inter-domain redundancy path computation methods based on PCE (Path Computation Element). Some inter-domain paths carry traffic that must be assured of high quality and high reliability transfer such as telephony over IP and premium virtual private networks (VPNs). It is, therefore, important to set inter-domain redundancy paths, i.e. primary and secondary paths. The first scheme utilizes an existing protocol and the basic PCE implementation. It does not need any extension or modification. In the second scheme, PCEs make a virtual shortest path tree (VSPT) considering the candidates of primary paths that have corresponding secondary paths. The goal is to reduce blocking probability; corresponding secondary paths may be found more often after a primary path is decided; no protocol extension is necessary. In the third scheme, PCEs make a VSPT considering all candidates of primary and secondary paths. Blocking probability is further decreased since all possible candidates are located, and the sum of primary and secondary path cost is reduced by choosing the pair with minimum cost among all path pairs. Numerical evaluations show that the second and third schemes offer only a few percent reduction in blocking probability and path pair total cost, while the overheads imposed by protocol revision and increase of the amount of calculation and information to be exchanged are large. This suggests that the first scheme, the most basic and simple one, is the best choice.

Obtaining current traffic matrices is essential to traffic engineering (TE) methods. Because it is difficult to monitor traffic matrices, several methods for estimating them from link loads have been proposed. The models used in these methods, however, are incorrect for some real networks. Thus, methods improving the accuracy of estimation by changing routes also have been proposed. However, existing methods for estimating the traffic matrix by changing routes can only capture long-term variations and cannot obtain current traffic matrices accurately. In this paper, we propose a method for estimating current traffic matrices that uses route changes introduced by a TE method. In this method, we first estimate the long-term variations of traffic by using the link loads monitored at previous times. Then, we adjust the estimated long-term variations so as to fit the current link loads. In addition, when the traffic variation trends change and the estimated long-term variations fail to match the current traffic, our method detects mismatch. Then, so as to capture the current traffic variations, the method re-estimates the long-term variations after removing monitored data corresponding to the end-to-end traffic causing the mismatches. We evaluate our method through simulation. The results show that our method can estimate current traffic matrices accurately even when some end-to-end traffic changes suddenly.

The Internet has grown extremely fast in the last two decades. The number of routes to be supported by the routers has become very large. Moreover, the number of messages exchanged to distribute the routes has increased even faster. In this paper, we propose SpliTable, a scalable way to support the Internet routes in a Service Provider network. In our proposal, BGP route selection is done by distributed servers on behalf of the routers. They are called route selection servers. The selected routes are then stored in distributed routing tables. Each router maintains only its share of Internet routes, not the routes for each Internet prefix as it is the case today. We adapted the concept of Distributed Hash Tables (DHT) for that purpose. We show analytically that our proposal is more scalable in the number of routes supported in each router than current iBGP route distribution solutions. Moreover, the number of control messages exchanged with our proposal is bounded contrary to current sparse iBGP route distribution solutions which may never converge. We confirm these findings in an evaluation of a prototype implementation.

In recent years, there has been a rapid growth in applications such as World Wide Web browsing, which are characterized by fairly short sessions that transfer substantial amounts of data. Conventional connection-oriented and datagram services are not ideally engineered to handle this kind of traffic. We present a new ATM service, called Dynaflow service, in which virtual circuits are created on a burst-by-burst basis and we evaluate key aspects of its performance. We compare Dynaflow to the Fast reservation protocol (FRP) and show that Dynaflow can achieve higher overall throughput due to the elimination of reservation delays, and through the use of shared "burst-stores. " We study the queueing performance of the dynaflow switch and quantify the relationship between the loss ratio and the buffer size.

This paper proposes the hierarchical cloud-router network (HCRN) to overcome the scalability limit in a multi-layer generalized multi-protocol label switching (GMPLS) network. We define a group of nodes as a virtual node, called the cloud-router (CR). A CR consists of several nodes or lower-level CRs. A CR is modeled as a multiple switching capability (SC) node when it includes more than one kind of SC, which is fiber SC, lambda SC, time-division multiplexing (TDM) SC, packet SC, even if there are no actual multiple switching capability nodes in the CR. The CR advertises its abstracted CR internal structure, which is abstracted link state information inside the CR. A large-scale, multi-layer network can then achieve scalability by advertising the CR internal structure throughout the whole network. In this scheme, the ends of a link connecting two CRs are defined as interfaces of the CRs. We adopt the CR internal cost scheme between CR interfaces to abstract the network. This CR internal cost is advertised outside the CR via the interfaces. Our performance evaluation has shown that HCRN can handle a larger number of nodes than a normal GMPLS network. It can also bear more frequent network topology changes than a normal GMPLS network.