摘要:
Technologies are generally described for substantially maximizing capacity in a wavelength division multiplexing (WDM) passive optical network (PON). An “achievable rate region” may be defined as a set containing admissible traffic rates of a given WDM PON system such that a volume of an achievable rate region is proportional to a capacity of the network. Deriving the achievable rate region for a particular network, decisions may be made whether incoming traffic rate can or cannot be achieved for that network. Moreover, the achievable rate region may be used to construct a WDM PON utilizing a minimum number of wavelengths, a minimum number of lasers with narrowest tuning ranges, and a minimum number of receivers, thereby reducing a capital expenditure in building the PON system.
摘要:
Technologies are generally described for network traffic scheduling in a wavelength division multiplexing (WDM) passive optical network (PON). Dynamic wavelength assignment and time allocation in hybrid WDM/TDM PONs with tunable lasers as optical light generators is accomplished by mapping the scheduling into a multi-processor scheduling problem with wavelength channels as machines and ONU requests as jobs. Wavelengths may be considered as parallel identical machines. Taking laser tuning time into consideration preemptive and non-preemptive scheduling with the objective of minimizing the latest job completion time is computed employing a number of heuristic algorithms. The algorithms compute two extreme cases of zero and infinity laser tuning time, respectively. Using the results from these two extreme cases, the heuristic scheduling schemes for the case of arbitrary laser tuning time yield close average latest job completion times for both schedule types.
摘要:
Techniques for partitioning and/or combining at least a portion of an optical network tree including one or more array waveguide gratings (“AWGs”) and fibers in wavelength division multiplexing (“WDM”) passive optical networks (“PON”) with cascaded AWGs are disclosed. Example methods include deriving the optimal positions for the AWGs to minimize fiber cost and then determining the arrangement of cascaded AWGs to minimize the total cost of AWGs and fibers. Determining the arrangement of cascaded AWGs may include recursive partitioning followed by recursive combination. An example recursive partition-combination based algorithm for optimizing a tradeoff between the AWG cost and the fiber cost is disclosed.
摘要:
A differential frame-based scheduling scheme is employed for input queued (IQ) switches with virtual output queues (VOQ). Differential scheduling adjusts previous scheduling based on a traffic difference in two consecutive frames. To guarantee quality of service (QoS) with low complexity, the adjustment first reserves some slots for each port pair in each frame, then releases surplus allocations and supplements deficit allocations according to a dichotomy order, designed for high throughput, low jitter, fairness, and low computational complexity.
摘要:
Techniques for partitioning and/or combining at least a portion of an optical network tree including one or more array waveguide gratings (“AWGs”) and fibers in wavelength division multiplexing (“WDM”) passive optical networks (“PON”) with cascaded AWGs are disclosed. Example methods include deriving the optimal positions for the AWGs to minimize fiber cost and then determining the arrangement of cascaded AWGs to minimize the total cost of AWGs and fibers. Determining the arrangement of cascaded AWGs may include recursive partitioning followed by recursive combination. An example recursive partition-combination based algorithm for optimizing a tradeoff between the AWG cost and the fiber cost is disclosed.
摘要:
Technologies are generally described for network traffic scheduling in a wavelength division multiplexing (WDM) passive optical network (PON). Dynamic wavelength assignment and time allocation in hybrid WDM/TDM PONs with tunable lasers as optical light generators is accomplished by mapping the scheduling into a multi-processor scheduling problem with wavelength channels as machines and ONU requests as jobs. Wavelengths may be considered as parallel identical machines. Taking laser tuning time into consideration preemptive and non-preemptive scheduling with the objective of minimizing the latest job completion time is computed employing a number of heuristic algorithms. The algorithms compute two extreme cases of zero and infinity laser tuning time, respectively. Using the results from these two extreme cases, the heuristic scheduling schemes for the case of arbitrary laser tuning time yield close average latest job completion times for both schedule types.
摘要:
A differential frame-based scheduling scheme is employed for input queued (IQ) switches with virtual output queues (VOQ). Differential scheduling adjusts previous scheduling based on a traffic difference in two consecutive frames. To guarantee quality of service (QoS) with low complexity, the adjustment first reserves some slots for each port pair in each frame, then releases surplus allocations and supplements deficit allocations according to a dichotomy order, designed for high throughput, low jitter, fairness, and low computational complexity.
摘要:
Technologies are generally described for substantially maximizing capacity in a wavelength division multiplexing (WDM) passive optical network (PON). An “achievable rate region” may be defined as a set containing admissible traffic rates of a given WDM PON system such that a volume of an achievable rate region is proportional to a capacity of the network. Deriving the achievable rate region for a particular network, decisions may be made whether incoming traffic rate can or cannot be achieved for that network. Moreover, the achievable rate region may be used to construct a WDM PON utilizing a minimum number of wavelengths, a minimum number of lasers with narrowest tuning ranges, and a minimum number of receivers, thereby reducing a capital expenditure in building the PON system.
摘要:
A method for delivering packets in a wireless communications system includes determining a cause of loss for a previously transmitted packet based on a packet acknowledgement corresponding to the previously transmitted packet, the packet acknowledgement including at least one of a wireless loss indicator and a congestion warning indicator for the wireless communications system. The method also includes adjusting a transmission parameter in a packet transmission protocol according to the cause of loss, and retransmitting the previously transmitted packet.
摘要:
Techniques for transferring data among nodes on a network are disclosed. Some example methods include a downloader-initiated random linear network coding algorithm. A downloading node may be aware of the chunks of original data held by neighboring nodes, and the downloading node can request linear combinations of chunks from the neighboring nodes that are linearly independent of any linear combinations of chunks already held by the downloading node.