Chapter 9
Emerging Hardware Technologies for IoT
Data Processing
Mahdi Nazm Bojnordi and Payman Behnam
No man has a good enough memory to be a successful liar.
Abraham Lincoln
Contents
9.1 Challenges for Data Processing in the Era of IoT ....................................... 434
9.1.1 IoT System Architecture......................................................... 434
9.1.2 Energy Efficiency as a Paramount Concern .................................... 435
9.1.3 Bandwidth Limitation for Big Data Processing ................................ 436
9.2 Recent Innovations for Bandwidth and Energy .......................................... 436
9.2.1 Heterogeneous Computing ...................................................... 436
9.2.2 In-Package Die Stacking ........................................................ 437
9.2.3 Emerging Memory Technologies ............................................... 438
9.2.4 Machine Learning Accelerators in the IoT Era ................................. 439
9.2.5 Approximate Computing ........................................................ 440
9.3 Near-Memory Processing ................................................................. 442
9.4 In Situ Processing for IoT Devices ....................................................... 443
9.4.1 Deep Binary Neural Network ................................................... 443
9.4.2 The MB-CNN Architecture ..................................................... 445
9.4.3 Memristive XNOR Convolution ................................................ 446
9.4.4 The MB-CNN Architecture ..................................................... 448
9.4.5 Potentials of the MB-CNN Accelerator......................................... 452
9.5 In Situ Data Clustering for IoT Servers .................................................. 453
9.5.1 Data Clustering .................................................................. 454
9.5.2 Applications of Data Clustering................................................. 454
9.5.3 Data Clustering with Rank-Order Filters ....................................... 456
9.5.4 Memristive k-Median Clustering ................................................ 457
9.5.5 MISC Building Blocks .......................................................... 459
9.5.6 Potentials of the MISC Accelerator ............................................. 465
References ........................................................................................ 466
M. N. Bojnordi () · P. Behnam
University of Utah, Salt Lake City, UT, USA
e-mail: bojnordi@cs.utah.edu
© Springer Nature Switzerland AG 2020
F. Firouzi et al. (eds.), Intelligent Internet of Things,
https://doi.org/10.1007/978-3-030-30367-9_9
433
Emerging Hardware Technologies for IoT
Data Processing
Mahdi Nazm Bojnordi and Payman Behnam
No man has a good enough memory to be a successful liar.
Abraham Lincoln
Contents
9.1 Challenges for Data Processing in the Era of IoT ....................................... 434
9.1.1 IoT System Architecture......................................................... 434
9.1.2 Energy Efficiency as a Paramount Concern .................................... 435
9.1.3 Bandwidth Limitation for Big Data Processing ................................ 436
9.2 Recent Innovations for Bandwidth and Energy .......................................... 436
9.2.1 Heterogeneous Computing ...................................................... 436
9.2.2 In-Package Die Stacking ........................................................ 437
9.2.3 Emerging Memory Technologies ............................................... 438
9.2.4 Machine Learning Accelerators in the IoT Era ................................. 439
9.2.5 Approximate Computing ........................................................ 440
9.3 Near-Memory Processing ................................................................. 442
9.4 In Situ Processing for IoT Devices ....................................................... 443
9.4.1 Deep Binary Neural Network ................................................... 443
9.4.2 The MB-CNN Architecture ..................................................... 445
9.4.3 Memristive XNOR Convolution ................................................ 446
9.4.4 The MB-CNN Architecture ..................................................... 448
9.4.5 Potentials of the MB-CNN Accelerator......................................... 452
9.5 In Situ Data Clustering for IoT Servers .................................................. 453
9.5.1 Data Clustering .................................................................. 454
9.5.2 Applications of Data Clustering................................................. 454
9.5.3 Data Clustering with Rank-Order Filters ....................................... 456
9.5.4 Memristive k-Median Clustering ................................................ 457
9.5.5 MISC Building Blocks .......................................................... 459
9.5.6 Potentials of the MISC Accelerator ............................................. 465
References ........................................................................................ 466
M. N. Bojnordi () · P. Behnam
University of Utah, Salt Lake City, UT, USA
e-mail: bojnordi@cs.utah.edu
© Springer Nature Switzerland AG 2020
F. Firouzi et al. (eds.), Intelligent Internet of Things,
https://doi.org/10.1007/978-3-030-30367-9_9
433