Needs for various delay-sensitive applications, such as Voice-over -IP (VoIP), Internet Protocol Television (IPTV), and interactive gaming, which requires tightly low end-to-end delay and delay jitter, and low packet loss rate have been increasing. In order to support this trend, it is essential to develop an efficient and QoS-aware networking technology. Optical Packet Switching (OPS) including Optical Burst Switching (OBS) is a networking technology enabling to switch and transport optical packet-level data traffic beyond simple point-to-point transmission. Especially, through OPS technology, we can achieve high multiplexing efficiency owing to packet-level switching as well as provide delay-guaranteed service owing to cut-through switching mechanism at core nodes. However, OPS technology has several limitations to support QoS-guaranteed services. First, it's not doable to provide absolute QoS even for high priority traffic such as delay-sensitive application traffic. Second, blocking rate among data bursts at core nodes is not quite acceptable to satisfy service requirements. Third, installation cost (i.e.,CAPEX) and operating cost (i.e., OPEX) could be increased by supplementary equipment such as Fiber Delay Line (FDL) buffer used for relaxing contention of data bursts. Lastly, network utilization of OPS is degraded as increasing traffic load because of high contention among data bursts.
In order to build QoS-guaranteed and cost-effective next generation packet network based on OPS, it is essential to provide revolutionary solutions for aforementioned problems of OPS. As a common solution for the problems of OPS, we considered flow control mechanism in OPS, naming OPS flow control. General concept of flow control is the process of managing the rate of data transmission between two nodes to prevent a fast sender from over running a slow receiver. A well-known mech-mechanism to know this congestion is to request acknowledged packet from destination node. However, this flow control in high layer (e.g.,TCP) cannot guarantee QoS requirements, such as delay, jitter, and loss of packet, of key services since it only controls the sending rate of source node and there is no way to know exact condition of overall network. Because of this reason, we studied flow control in OPS in order not only to guarantee QoS requirements but also to reduce CAPEX and OPEX cost by avoiding usage of FDL buffer.
This dissertation, first, introduces an OPS flow control scheme utilizing loss rate information in core nodes in order to guarantee QoS requirements, especially for delay-sensitive application traffic such as VoIP and IPTV. Second, this dissertation presents an OPS flow control scheme utilizing traffic load information in core nodes, so that it traces optimal performance in OPS network, having lowest blocking rate on fluctuating network traffic loads. This dissertation lastly addresses an OPS flow control scheme supporting loss-free slotted OPS by real-locating number of slots and slot position in synchronized slot series according to ever-changing network traffic condition.
This dissertation mainly studies aforementioned OPS flow control schemes. First, we study an OPS flow control scheme utilizing loss rate information in core nodes, called it as Adaptive Loss-aware Flow Control (ALFC) scheme, in order to guarantee QoS requirements, especially for delay-sensitive application traffic such as VoIP and IPTV. ALFC scheme utilizes an adaptive off set time to allow a high priority for delay-sensitive application traffic based on the loss-rate information generated at the core nodes. Through the management of delay and jitter factors, the ALFC scheme controls the upper-bounds of delay and jitter, so that it also guarantees the delay and jitter requirements of delay-sensitive application traffic.
Second, we study an OPS flow control scheme utilizing traffic load information in core nodes, called it as Adaptive Load-aware Burst Assembly (ALBA) scheme. The ALBA scheme adjusts burst size optimized to current traffic load, so that FDL buffers can accomplish the best performance without changing the granularity. The advantage of the proposed ALBA scheme adjusting the burst size by monitoring the network traffic-load variation is that it is not necessary for FDL buffers to transform the hardware system instantly for achieving the best performance.
Third, we study an OPS flow control scheme supporting loss-free slotted OPS by reallocating number of slots and slot position in OPS superframe, which is a cyclic period consisting of multiple time-slots, on fluctuating network traffic condition. Time-slots for each source node are basically allocated by proposing Tree-based Slot Allocation(TSA) algorithm, which allocates more time-slots than average necessity to source nodes far from destination node and allocates fewer time-slots than average necessity to source nodes closed to destination node through applying different weight in the progress of time-slot allocation. Since source nodes closed to destination node have more chances to utilize empty time-slots, allocating more time-slots to source node far from destination node can enhance network utilization. In order to manage time-slots efficiently, we propose OPS superframe consisting of multiple time-slots transmitted by source node toward same destination node. Multiplexing optimization among OPS superframe radically reduces the number of wavelength consumption, so that it makes slotted OPS with TSA algorithm as a cost-effective next generation packet network. However, since slot allocation works are completed at the step of network planning, it's not affordable to change number of time-slots and slot position in OPS superframe according to changed traffic load condition per source node. Therefore, an OPS flow control scheme is essential to build loss-free OPS network. When incoming traffic load at source node increases much more than expectation, source node requests additional time-slots to control node by following proposing flow control procedure. Likewise, when incoming traffic load at source node is reduced, source node returns additionally allocated time-slots to control node through proposing flow control procedure.
A major contribution of this dissertation is to suggest OPS flow control schemes providing QoS-guaranteed services as well as achieving high network utilization on fluctuating network traffic condition. Proposed QoS-aware flow control schemes introduced in this dissertation are expected to be utilized by network operators who develop and operate cost-efficient and revenue generating network. These are also expected to be used a reference for researchers and network venders who design and develop next generation packet network.
