This paper proposes a simple EPC queueing model to evaluate the network performance of the control plane in LTE core networks, and to clarify the impact of the parameter for the radio interface, the inactivity timer, on the performance of core networks. The proposed EPC queueing model incorporates the protocol behaviors of state transition in session management, and introduces a parameter, the mixture ratio of batch arrival (MRBA), to represent burst traffic or spike traffic in the input of the queueing model. The network performance is evaluated in terms of loss rate and latency for connection setup. Numerical results show that MRBA substantially affects the performance of a core network. It is also revealed that some values of the inactivity timer locally degrade the performance of a core network depending on the specific conditions of the input traffic.

Content Delivery Network services, which were originally aimed at achieving fast content delivery, began being used for fast delivery of data stored in cloud services, but they are only effective for frequently accessed data. In this paper, we present an overlay architecture of inter- data center (DC) networking for fast content delivery that is effective for long-tail contents. Based on round-trip time measurements between DCs, content distribution trees (CDTs) are dynamically reconfigured to maximize data transfer throughputs from origin to all other DCs. Furthermore, to enhance throughput by using parallel data transfer on overlay paths (OPs) on a current CDT, a content file are segmented off-line into blocks that are then named at a content server so that each of multiple parallel block-streams can be transferred by an assigned HTTP proxy server at each hop of the OP. Numerical examples obtained by using multiple traceroute servers located in an Asia-Pacific region show that throughput estimates of optimized OPs between some country pair are more than ten times greater than that of direct paths.

Traffic engineering aims to optimize the utilization of existing network resources for load balance and failure recovery, and these are to be accomplished in a scalable fashion. This paper proposes a traffic engineering scheme using multiple multipoint-to-point (m-t-p) label switched paths (LSP) which can reduce the number of LSP and required labels in links. The scheme consists of m-t-p LSP creation and flow assignment. Routes are first selected, and m-t-p LSP are designed to include them. The m-t-p LSP design problem is formulated as a 0-1 integer programming problem. The flow assignment problem is formulated as a mixed integer programming problem in which maximum link load, i.e., maximum congestion, is minimized. Numerical comparisons with the conventional point-to-point LSP approach show that the m-t-p LSP approach can reduce the number of required LSP and labels. Moreover, numerical comparisons with conventional shortest path fast-based flow assignment show that our flow assignment scheme can reduce maximum link load.

To develop survivable and cost-efficient networks, it is important to consider practical restoration control in network design. We propose a design for joint capacity assignment to state-independent working paths and state-independent spare paths. The assignment is formulated as an integer programming problem. The use of state-independent paths simplifies the restoration process and shortens the restoration time, and our design is especially suitable to pre-planned distributed restoration schemes. We further apply it to a strategy in which all paths are state-dependent, and to one whose spare paths are state-dependent. A numerical comparison of capacity costs, in which our design was applied to specific network examples, shows that costs for a state-independent approach are only slightly higher than those for state-dependent approach.

QoS restoration, the capability of a network to help maintain the QoS above pre-defined minimum requirements in case of failures or load fluctuations, is proposed. If failures or load fluctuations occur, a network with QoS restoration reallocates surplus network resources from working connections as well as from pre-assigned spare capacity by using a distributed algorithm to restore the degraded connections. This distributed algorithm is composed of the flooding of failure or load fluctuation notification, working connection reconfiguration signaling, and a conventional self-healing or distributed algorithm against load fluctuations. Theoretical maximum restoration ratios for various failures were calculated by integer programming. These analytical evaluations show that resource reallocation allows high ratio of restoration from single-link failures having no spare resources, and double-link failures having spare resources prepared for single-link failures. Simulation results show that the proposed distributed algorithm achieves a restoration ratio close to the theoretical maximum regardless of whether pre-assigned spare capacity exists.

This paper proposes an optimal design scheme for survivable wavelength-division multiplexing (WDM) transport networks in which fast restoration can be achieved by using predetermined restoration paths that are independent of failure locations and which have one to one correspondence with working paths. Integer programming-based design problems are formulated to optimally determine working and their corresponding restoration paths, the number of fibers in each span, and the number of optical cross connects (OXCs) in each node. In these optimization problems, total facility cost, which includes not only transmission cost but also cross connection, is minimized. This design scheme can handle the dedicated/shared allocation of spare resources and several parameters, such as the maximum available number of wavelengths per fiber /spl Omega/, the matrix size of each OXC, and cross-connection to transmission cost-coefficient ratio /spl gamma/. Total fiber length, the total number of OXCs, etc., are evaluated from obtained design results for these options and parameters. Numerical examples show that the dependency of total fiber length and the total number of OXCs on /spl gamma/ is relatively small, and cross-connection to transmission cost ratios (/spl gamma/ times the total number of OXCs divided by total fiber length) in the dedicated and shared cases are almost the same for each combination of /spl gamma/ and /spl Omega/.