We present a high-throughput FPGA design for supporting high-performance network switching. FPGAs have recently been attracting attention for datacenter computing due to their increasing transceiver count and capabili...
详细信息
ISBN:
(纸本)9781450370998
We present a high-throughput FPGA design for supporting high-performance network switching. FPGAs have recently been attracting attention for datacenter computing due to their increasing transceiver count and capabilities, which also benefit the implementation and refinement of network switches. Our solution replaces the crossbar in favour of a novel, more pipeline-friendly approach, the "Combined parallel round-robin arbiter". It also removes the overhead of incorporating an often-iterative scheduling or matching algorithm, which sometimes tries to fit too many steps in a single or a few FPGA cycles. The result is a network switch implementation on FPGAs operating at a high frequency and with a low port-to-port latency. It also provides a wiser buffer memory utilisation than traditional Virtual Output Queue (VOQ)-based switches and is able to keep 100% throughput for a wider range of traffic patterns using a fraction of the buffer memory and shorter packets.
In a mobile computing or shared I/O environment, one data service can satisfy multiple requests. The data broadcast problem is one example of data service that has been extensively studied. Previous work in data broad...
详细信息
ISBN:
(纸本)1601320841
In a mobile computing or shared I/O environment, one data service can satisfy multiple requests. The data broadcast problem is one example of data service that has been extensively studied. Previous work in data broadcast typically focuses on how to shorten the average request waiting time or how to broadcast the data items to serve as many requests as possible in a limited time, but does not consider how to meet the deadlines of requests. In addition, such work typically considers individual data requests, as opposed to the multiple data requests associated with a transaction. In this paper, we discuss how to serve data in order to complete more transactions with multiple data requests before their deadlines. We assume that transactions can read one or more data items and the data access sequence does not matter. We propose a new data server scheduling strategy for read-only transactions with deadlines that considers the importance of transactions, and we introduce a new cost model to measure the system performance. The simulation results show that our novel algorithm achieves better performance for transactions with deadlines than existing scheduling algorithms.
With the advent of SMT multicore (Simultaneous Multithreading) processors co-scheduling of jobs on the same core, thus sharing core resources like L1, L2 caches there results a per process performance degradation depe...
详细信息
The proliferation of multi-core architectures has led to explosive development of parallel applications using programming models, such as OpenMP, TBB, and Cilk, etc. With increasing number of cores, however, it become...
详细信息
Wireless sensor networks (WSN) are designed for data gathering, processing and transmitting with particular requirements: low hardware complexity, low energy consumption, special traffic pattern support, scalability, ...
详细信息
Wireless sensor networks (WSN) are designed for data gathering, processing and transmitting with particular requirements: low hardware complexity, low energy consumption, special traffic pattern support, scalability, and in some cases, real-time operation. The emergence of wireless cyber physical systems leads to the need of real time scheduling of data packets. Developing packet scheduling algorithms in WSNs can efficiently enhance delivery of packets through wireless links. Packet scheduling is a process defined as selecting or rejecting a packet depending upon a decision. The packets are transmitted based on various algorithms within the network and there is a possibility of dropping the packets due to packet size, bandwidth, packet arrival rate, deadline of packet. To achieve real time delivery, the paths must deliver the data in time. Some of the algorithms have been selected for packet scheduling of real time data to achieve predictable and bounded end-to-end latencies while meeting the deadlines of queries, in which NJNC (Nearest Job Next-with combination) outperforms in mobility assisted data collection with combination of multiple requests served together in on-demand manner without starvation problem as in the case of existing schemes like first-come-first-serve (FCFS), shortest-job-next (SJN). The results shows that NJNC provides better performance than the nearest-job-next(NJN).
In modern embedded platforms, safety-critical functionalities that must be certified correct to very high levels of assurance may co-exist with less critical software that are not subject to certification requirements...
详细信息
In modern embedded platforms, safety-critical functionalities that must be certified correct to very high levels of assurance may co-exist with less critical software that are not subject to certification requirements. Upon such platforms one seeks to satisfy two, sometimes contradictory, goals: (i) being able to certify the safety-critical functionalities under very conservative assumptions, and (ii) ensuring high utilization of platform resources even when actual run-time behavior does not live up to such pessimistic expectations. This paper describes efforts at designing scheduling algorithms that balance these two requirements, when scheduling recurrent tasks that are triggered by external events of unknown exact frequency.
With the increasingly serious energy and environmental problems, renewable energy has been favored all over the world. In order to ensure the safe and stable operation of the power grid when wind power and photovoltai...
详细信息
ISBN:
(纸本)9781665491129
With the increasingly serious energy and environmental problems, renewable energy has been favored all over the world. In order to ensure the safe and stable operation of the power grid when wind power and photovoltaic are connected to the grid, the power system often takes some technical means and measures to absorb the output from wind power and photovoltaic. This paper mainly studies the smart grid scheduling algorithm based on artificial neural network. In this paper, a BP neural network model based on load prediction is constructed. By integrating the learning network rate, elastic method is adopted to modify and optimize the BP neural network model. The load data obtained from the modified model is used for power grid dispatching. The experimental results show that the BP neural network is far superior to the traditional dispatching model in load prediction and power grid dispatching.
Addressing the requirements of Industrial Internet of Things (IIoT) in Industry 4.0, the Time Slotted Channel Hopping (TSCH) protocol of the IEEE 802.15.4e amendment has been proposed. However, the lack of a defined s...
详细信息
ISBN:
(数字)9798350311617
ISBN:
(纸本)9798350311624
Addressing the requirements of Industrial Internet of Things (IIoT) in Industry 4.0, the Time Slotted Channel Hopping (TSCH) protocol of the IEEE 802.15.4e amendment has been proposed. However, the lack of a defined scheduling procedure in the standard remains an open research area. Existing reinforcement learning-based scheduling proposals demonstrate great potential for this technique due to the ongoing observations within the network environment. Beneficial for real-world scenarios where network conditions are volatile and unpredictable. This work presents QL- TSCH-plus, an enhancement of the existing QL- TSCH scheduler that reduces energy consumption by adapting the Action Peeking mechanism to a distributed scheme. Instead of continuously listening to neighboring nodes communication, QL- TSCH-plus allows nodes to broadcast the learned transmission slots for updating the Action Peeking Tables and allocating reception slots, reducing energy use by up to 47% compared to QL-TSCH. This novel approach also maintains reliability and timeliness, demonstrating significant potential for efficient scheduling in TSCH networks, making it suitable for the IIoT.
We present an adaptation of the dual priority scheduling algorithm to schedule both hard real-time periodic tasks and soft-aperiodic tasks in shared memory multiprocessor systems. The goal is to achieve low mean aperi...
详细信息
Advance resource reservations allow clients to have exclusive access to Grid resources. The goal is to deliver higher levels of Quality of Service. However, in practice, the support for advance reservations in Grids h...
详细信息
暂无评论