China
Africa
Explore chapters and articles related to this topic
Localization Protocols for Wireless Sensor Networks
Published in Mohammed Usman, Mohd Wajid, Mohd Dilshad Ansari, Enabling Technologies for Next Generation Wireless Communications, 2020
Ash Mohammad Abbas, Hamzah Ali Abdul Rahman Qasem
A localization scheme called Connectivity-Based and Anchor Free Three Dimensional Localization (CATL) is proposed in Tan et al. (2013) for large-scale WSNs with concave regions. CATL is based on the identification of special nodes called notches, where end-to-end shortest paths between the source and the destination bend. Also, at notches a significant difference between the Euclidean distance and the hop-count-based distance is started to be observed. CATL consists of an iterative algorithm using multilateration-based localization to avoid notches. The method of identifying notch nodes in the network is based on the intuition that in the global shortest path tree, such type of nodes often possess a subtree that is relatively fatter as compared to ordinary nodes. To identify notches, one needs to detect the fat-tree abnormality in the shortest path tree. A major feature of CATL is that it is independent of the network boundaries.
On-Chip Bus vs. Network-on-Chip
Published in Marcello Coppola, Miltos D. Grammatikakis, Riccardo Locatelli, Giuseppe Maruccia, Lorenzo Pieralisi, Design of Cost-Efficient Interconnect Processing Units, 2020
Marcello Coppola, Miltos D. Grammatikakis, Riccardo Locatelli, Giuseppe Maruccia, Lorenzo Pieralisi
In the wormhole-routed Butterfly Fat Tree NoC topology (BFT), the layout is a fat tree [263]. Each tree node is represented with a pair of coordinates: its level in the tree (starting from zero at the leaves) and its position, e.g. in right to left ordering. Tree leaves correspond to IP or processor cores. Each switch at layer one is connected to four such blocks. Intermediate routers (at layer two and above) are allocated two parent ports, and four child ports, or connections, i.e. BFT is a 4-ary fat tree. Notice that the number of routers at each level decreases by a factor of two. Thus, BFT provides an overall cores to routers ratio of 2:1, compared to 4:1 for mesh. However, this architecture enables simple traffic aggregation to/from a particular set of cores and regular structuring of the switches in the layout, simplifying design. In fact, BFT trades throughput for reducing area overhead and power efficiency, more than the SPIN. For a more precise definition of general fat tree topologies refer to Section 3.5.
Parallel Architectures
Published in Pranabananda Chakraborty, Computer Organisation and Architecture, 2020
The drawbacks of a (binary) tree having the possibility of a bottleneck at the upper levels have been reduced by way of increasing the number of links in the upper levels of a tree. The modified tree is known as fat tree, as introduced by Leiserson in 1985 in which each node in the tree (except at the top level) has more than one parent. Two types of fat tree are shown in Figure 10.15d and e that indicates the channel width of a fat tree increases as one traverse from leaves to the root. The fat tree structure-wise resembles a natural tree in which branches get thicker towards the root. A fat tree structure has been applied in Connection MachineCM-5 from Thinking Machine Corporation replacing the previous hypercube implementation which was used in its earlier version CM-2.
Dynamic reliability modeling of cyber-physical edge computing network
Published in International Journal of Computers and Applications, 2021
Data-centers are used in CPS. The various classifications of data-centers for CPS has been summarized in [26]. A very common structure is the Fat-tree data center networks for CPS [27,28]. Most Fat-tree networks use global round robin (GRR) routing algorithm for transversing packets/IO streams between upper-bound and lower bound switch/access point sources. According to [27], the GRR offers best delay/throughput performance. There is Fat-tree network (either type 1 or 2) that uses explicit route construction for its global packing number considering all the supported hosts [29]. Their constructed routes seem to be load-balanced and require minimal link capacities. In the case of Fat-tree-based on-chip network [30], it has been argued that it has several advantages over traditional mesh or torus-based networks especially in terms of throughput, power efficiency, and latency.