- Towards 3D Manycore-Based-Noc Architecture In Two Layers Of Processors And Memories For A Better Congestion
Gaffour Khadidja1; MK Benhaoua1, 2; T Djeradi3; AEH Benyamina1; P Boulet4 1Computer Science Department, LAPECI, University of Oran 1 BP 1524 ELM_Naouer 31000, Oran, Algeria2University of Mascara, Mascara, Algeria 3 Schlumberger Integrated Solutions, Houston, USA.4 University Lille1, CRISTAL, CNRS, UMR 8022 F-59650Villeneuve d’Ascq, France
The next generation of manycore architecture with hundreds of cores will replace soon the current multicore architecture. If we double the number of cores per chip every year, we will reach a thousand in the upcoming few years. In this matter, the Network-on-Chip (NoC) is proposed as alternative pattern which is considered as an emerging technology for distributed embedded systems. A novel 3D architecture (3DPM) on two layers, the first layer is composed of different types of processors and the second layer is composed of different types of memories. The 3DPM is a good alternative to cope with the limitations of 2D manycore architecture. The main problem in 2D NoC architecture is due to the congestion when transmitting the big data packets from the shared memory to all the processors how execute different applications tasks. However, the communication delay and power consumption of global interconnections become the major bottleneck. In this work we combined with two tools GEM5 and Noxim to realize this novel architecture. We are planning to compare a few mapping algorithms in both 2D and 3DPM architectures.
- Monthly Rainfall Prediction Using Backpropagation Algorithm
Melissa P. Juanillo, Dr. Ma. Christina R. Aragon, Dr. Rosmina Joy M. Cabauatan and Elsa V. Isip Computer Science Department, College of Information Technology Education, Technological Institute of the Philippines, Quezon City, Philippines
Rainfall affects the planning and management of water resources, traffic, flood, sewer systems, crop production, and other human activities. Predicting the rainfall is a challenging task and highly complicated due to the complexity of the physical processes influencing the occurrence of rainfall. A predictive model is developed using Artificial Neural Network (ANN) and implemented the Backpropagation algorithm for monthly rainfall prediction in Metro Manila. The data obtained from Philippine Atmospheric, Geophysical and Astronomical Services Administration (PAGASA) comprised 35 years of actual and forecast rainfall from Metro Manila station. The ANN is modelled with 34 years of data, 80% for training, 20% for validation and testing, and evaluated with the remaining 1 year. Evaluation and comparison of predictive performances between ANN and PAGASA were performed which reveals that ANN model provides promising results in approximating monthly rainfall patterns and can be used as an appropriate method for predicting the monthly rainfall.
- Multiplexing Voip Packets Over Internet Telephony Transport Protocol (ITTP)
Mosleh M. Abualhaj1, Sumaya N. Al-khatib2 and Mahmoud Baklizi3 1Department of Networks and Information Security, Al-Ahliyya Amman University, Amman, Jordan2Department of Software Engineering, Al-Ahliyya Amman University, Amman, Jordan3Department of Computer Networks Systems, the World Islamic Sciences & Education University W.I.S.E, Amman, Jordan
Voice over IP (VoIP) is a technology of making phone calls over IP network. A considerable number of VoIP applications have emerged in the last decade, to make calls over the Internet. Accordingly, a huge number of VoIP packets are running over the Internet, which consume a considerable share of bandwidth. VoIP applications produce packets with small payload, which cause a considerable header overhead. Whereas, the header overhead is between around 46.4% to 72.2%, depends of the payload size, when using Internet Telephony Transport Protocol (ITTP)/IP protocol. VoIP packets header multiplexing is one of the most common techniques that used to reduce header overhead. In this paper, we propose a new multiplexing technique, called Delta-Multiplexing over ITTP (D-Mux-ITTP). The D-Mux-ITTP is aiming to reduce the wasted bandwidth resulting from the 26-byte ITTP/IP VoIP packet header, by combining the packets destined to the same destination into a single ITTP/IP header.