Cloud computing is among the most rapidly advancing domains and computing paradigms due to its significant impact on the advancement of technology, scientific research, the business market, and its influence on social life. Cloud services have become the computational model for many related disciplines, such as data science and machine learning. Moreover, the adoption of cloud computing services by industries and individuals is generating increased demand for industry skills in this domain. To address this expansion, higher education institutions have recognized the importance of incorporating cloud computing courses into their engineering curricula. This paper presents the development of a new Cloud Computing and Big Data multidisciplinary course at the Department of Electrical and Computer Engineering at Rutgers University in order to successfully train students in the competitive area of cloud computing. This paper will elaborate on, and provide rationale for the selection of course content, topics, tools, and the chosen pedagogical approach from a vast array of instructional materials, all aimed at delivering this course. Our aim has been to provide students with a strong foundation and self-initiative in this area so that even after the end of the course, they are fully capable of educating themselves and progressing in the demanding and competitive area of cloud computing. The course is unique in its interdisciplinary goals combining fields together such as: distributed and parallel computing, mathematics, programming and programming models, systems design, data science, and machine learning. Samples of students' projects and presentations that have resulted from the course offering will be presented. Moreover, we will further analyze and elaborate the “year by year” fine-tuning and adaptation of the material, tools, and teaching strategies used, based on our learning experiences as instructors, on our successful and unsuccessful strategies, and on the feedback we have received from students, from colleagues, and from industry employers.

While training students in electric circuits is difficult enough, training them in the art of design presents an entirely different challenge. Traditional methods used to teach theory and recipe-style laboratory experiments are ineffective at developing design skills. In this paper, we describe our approach at addressing this problem by incorporating a genuine design component in an introductory circuits lab, through a level-appropriate design challenge. The challenge is open-ended enough that a modest amount of searching, application of theory, and synthesis of knowledge from prior experiments is required to complete it. The task was to build a constant-current power source using an Arduino microcontroller. We found that students struggled to disassociate from the memorization-based learning approach they have become accustomed to, highlighting the need for more design-related content as early in the curriculum as possible. Despite the challenge, students generally found the design experiment engaging, as indicated by their responses to specific report questions. The paper also presents the results of using this design lab for ABET assessment of several student outcomes.

The performance of noncoherent M-ary frequency-shift keying (MFSK) with dual-branch postdetection switch and stay combining (SSC) in two-wave with diffuse power (TWDP) fading is studied. A closed-form expression for the average bit error rate (BER) of noncoherent MFSK with dual-branch postdetection SSC is derived. Monte Carlo simulation results are presented to verify the validity of the obtained analytical expressions. The effects of TWDP fading parameters on the relative performances of postdetection and predetection SSC receivers are studied. It is shown that postdetection SSC has superior performance over predetection SSC.

The performance of noncoherent M-ary frequency-shift keying (MFSK) with L branch signal-plus-noise (S+N) selection combining (SC) in two wave with diffuse power (TWDP) fading is studied. The fading on the diversity branches are assumed to be independent but not necessarily identically distributed. Two S+N SC receiver structures are examined and for each receiver a tractable bit error rate expression is derived. The effects of fading parameters on the relative performances of S+N SC and classical SC receivers are studied. Monte Carlo simulation results are presented to validate the obtained theoretical results.

The performance of noncoherent binary frequency shift keying (BFSK) with dual-branch postdetection switch-and-stay combining (SSC) in two-wave with diffuse power (TWDP) fading is studied. A closed- form expression for the average bit error rate (BER) of noncoherent BFSK with dual-branch postdetection SSC is derived. Previously obtained results on the performance of predetection SSC in TWDP fading are corrected. The performance of postdetection SSC is compared to the performance of predetection SSC and it is shown that postdetection SSC outperforms predetection SSC. The effects of the fading parameters on the relative performances of the two SSC receivers are studied. Monte Carlo simulation results are presented to verify the validity of the obtained analytical results.

The performance of a dual-branch decorrelator receiver operating in correlated Rayleigh and Rician fading channels in conjunction with selection combining (SC) and switch-and-stay combining (SSC) is analyzed and compared to the performance of a conventional SC and SSC receiver. For each combining scheme, analytical expressions for the average bit error rate (BER) of several modulation techniques of practical interest, the mean output signal-to-noise ratio (SNR) and the outage probability are derived. It is shown that the decorrelator receiver has superior performance over the conventional receiver by as much as 2.83 dB in average SNR. The effects of modulation order, correlation coefficient and the severity of fading on the performance of the decorrelator receiver are studied.

The performance of a dual-branch decorrelator receiver operating in correlated Rayleigh and Rician fading channels in conjunction with selection combining (SC), square- law combining (SLC) and equal gain combining (EGC) diversity is analyzed and compared to the performance of a conventional SC, SLC and EGC diversity receiver. Analytical expressions for the average symbol error rate (SER) and the average bit error rate (BER) of several modulation techniques of practical interest, the mean output signal-to-noise ratio (SNR) and the outage probability are derived. It is shown that the decorrelator receiver has superior performance over the conventional receiver by as much as 2.1 dB in average SNR when SC is employed. Interestingly, it is also shown that the mean output SNR of the decorrelator SC receiver improves with increasing the correlation coefficient while that of the conventional SC receiver degrades with increasing the correlation coefficient. In Rician fading, the results indicate that the outage probability of the decorrelator receiver is much less than the outage probability of the conventional receiver and the gap between the two increases as the channels become less faded.

The performance of a dual-branch decorrelator receiver operating in correlated Rayleigh and Rician fading channels in conjunction with switch-and-stay combining (SSC) is analyzed and compared to the performance of a conventional SSC receiver. Analytical expressions for the average bit error rate (BER) of several modulation techniques of practical interest, the mean output signal-to-noise ratio (SNR) and the outage probability are derived. It is shown that the decorrelator receiver has superior performance over the conventional receiver by as much as 2.83 dB in average SNR. The effects of modulation order, correlation coefficient and the severity of fading on the performance of the decorrelator receiver are studied.

The performances of two signal-plus-noise (S + N) selection diversity receivers employing noncoherent M-ary frequency shift keying (MFSK) in slow, flat, correlated Rayleigh and Rician fading channels are examined. The branches are assumed to be equally correlated for Rician fading. The correlation model for Rayleigh fading is more general than the equally correlated scenario and includes it as a special case. Analytical expressions are derived for the average symbol error rate (SER) and the average bit error rate (BER) of each receiver structure. Extensive Monte Carlo simulation results are presented to validate the analytical expressions. The performances of the S + N SC receivers are compared to the performance of the classical SC receiver. The effects of correlation, average fading power imbalance and diversity order on the performances of the S + N SC receivers are examined.

The performance ofM-ary frequency shift keying (MFSK) with dual-branch postdetection switch-and-stay combining (SSC) in correlated Rayleigh and Rician fading is studied. Closed-form expressions are derived for the average bit error rate (BER) of MFSK with postdetection SSC in correlated Rayleigh and Rician fading. Optimum switching thresholds that minimize the average BER are obtained. Extensive Monte Carlo simulations are presented to test the validity of the analytical results. The performance of MFSK with dual-branch postdetection SSC is compared to the performance of MFSK with dual-branch predetection SSC. The effects of correlation, fading parameter, fading power imbalance and switching threshold on the relative performance of the postdetection SSC compared to predetection SSC are examined.

The performances of two signal-plus-noise (S + N) selection diversity receivers employing M-ary noncoherent frequency shift keying (NCFSK) in slow, flat, correlated Rayleigh fading channels are examined. The correlation model is more general than the equally correlated scenario and includes it as a special case. Analytical expressions are derived for the average symbol error rate (SER) and the average bit error rate (BER) of each receiver structure. Extensive Monte Carlo simulation results are presented to validate the analytical expressions. The performances of the S + N SC receivers are compared to the performance of the classical SC receiver. The effects of correlation, average fading power imbalance and diversity order on the performances of the S + N SC receivers are examined.

The performance of binary frequency-shift keying (BFSK) and M-ary frequency shift keying (MFSK) with dual-branch postdetection switch-and-stay combining (SSC) in correlated Rayleigh and Rician fading is studied. Two postdetection SSC receivers are considered and the performances of noncoherent BFSK and MFSK are analyzed. Closed-form expressions are derived for the average bit-error rate (BER) of BFSK and MFSK with postdetection SSC in correlated Rayleigh and Rician fading. Optimum switching thresholds that minimize the average BER are obtained. Extensive Monte Carlo simulations are presented to test the validity of the analytical results. The performance of dual-branch postdetection SSC is compared with the performance of dual-branch predetection SSC. The effects of correlation, fading parameter, average fading power imbalance, and switching threshold on the performances of postdetection SSC receivers are examined

The performance of dual-branch predetection switch-and-stay combining (SSC) in correlated Rician fading is considered in conjunction with several modulation formats. Analytical expressions are derived for the average symbol error rate (SER) of predetection SSC in correlated Rician fading. Switching thresholds that minimize the average SER are obtained. The impact of fading factor and the fading correlation on the performance of predetection SSC is studied. Analytical expressions for the output average signal-to-noise ratio (SNR) and the outage probability are derived. The validity of the analytical expressions are verified using Monte Carlo simulation.

The performance of binary frequency-shift keying (BFSK) with dual-branch postdetection switch and stay combining (SSC) in correlated Rayleigh fading is considered. A closedform expression for the average bit error rate (BER) of BFSK with dual-branch postdetection SSC is derived. Optimum switching thresholds that minimize the average BER are obtained. It is shown that, for a fixed value of average signal-to-noise ratio (SNR), the optimum switching threshold is a decreasing function of branch correlation. Our analytical results are validated using Monte Carlo simulation results. The performance of dual-branch postdetection SSC is compared to the performance of dual-branch predetection SSC in correlated Rayleigh fading and it is shown that postdetection SSC has superior performance over predetection SSC for all values of SNR.

The performance of dual-diversity predetection switch-and-stay combining (SSC) in correlated Rician fading is considered in conjunction with several modulation formats. Analytical expressions are derived for the average symbol error rate (SER) of predetection SSC in correlated Rician fading. Switching thresholds that minimize the average SER are obtained. The impacts of fading parameter and the fading correlation on the performance of predetection SSC are studied. Analytical expressions for the output average signal-to-noise ratio (SNR) and the outage probability are derived. The validity of our analytical expressions are verified using Monte Carlo simulations.

The performances of general M-ary and binary orthogonal noncoherent frequency-shift keying signal-plus-noise (S+N) receiver structures in slow, flat Rician fading channels are examined. The fading on the diversity branches is assumed to be independent, but not necessarily identically distributed. For each receiver, a symbol-error probability expression is derived as a single integral with finite integration limits. Extensive Monte Carlo simulation results are presented to validate the analytical expressions. The results indicate that, unlike previously published results, the performance of S+N selection combining (SC) is not always superior to that of classical SC in fading and is dependent on the value of the signal-to-noise ratio (SNR), the modulation order, the diversity order, the multipath intensity profile, and the fading parameter. It is further shown that increasing the number of diversity branches in a S+N SC receiver does not necessarily decrease the probability of error if the system is operating in the low-SNR region. The performances of S+N SC schemes are also compared with the performances of equal-gain combining and square-law combining receivers.

The performances of dual-branch postdetection switch-and-stay combining (SSC) for noncoherent orthogonal binary frequency-shift keying and noncoherent orthogonal M-ary frequency-shift keying operating over slow, flat Rician fading channels are a topic of current interest. It is shown that recently published analyses obtain incorrect final results because some incorrect probability distribution parameters are used in the analyses. The previous results are corrected, and it is seen that the final results change significantly. The revised results are used to obtain optimum switching thresholds that minimize the average bit-error rate (BER) performance of these systems. Monte Carlo simulations are presented to verify the results, and it is shown that for a given BER, the signal-to-noise ratio difference of postdetection SSC and predetection SSC is much less than previously reported

The performance of M-ary orthogonal noncoherent frequency-shift keying (NCFSK) with N branch signal-plus-noise (S + N) selection combining (SC) in Nakagami-m fading (m, integer) is studied. Both independent, identically distributed (i.i.d) and independent, nonidentically distributed (i.n.d) diversity branches are considered and two S + N SC receiver structures are examined. The performances of the S + N SC receivers are compared to those of classical SC and square-law combining (SLC) receivers. The effects of modulation order, fading parameter and the number of diversity branches on the performance of S + N SC are compared to the effects on the performances of classical SC and SLC. For example, it is shown that in an i.n.d fading channel, the value of signal-to-noise ratio (SNR) at which the error rate curves of classical SC and S + N SC cross, decreases as the modulation order, M, increases. Our results indicate that in i.n.d fading channels classical SC outperforms S + N SC for small ranges of SNR, while for moderate to large SNR values S + N SC has superior performance over classical SC. It is also shown that increasing the diversity order will increase the performance gap of S 4N SC over classical SC and over SLC in both i.i.d and i.n.d Nakagami-m fading channels

In hybrid selection/maximal ratio combining (H-S/MRC) the receiver selects the L branches with the largest signal-to-noise ratios (SNRs) from N available diversity branches and performs maximal ratio combining (MRC). A penalty is incurred with respect to MRC which can be defined as the increase in SNR required for H-S/MRC to achieve the same symbol error probability as MRC. Recently, this penalty was investigated specifically for two-dimensional modulations. In this paper, we derive the asymptotic SNR penalties in slow Rayleigh fading for small and large values of SNR for general uncoded signaling constellations.

The performance of M-ary orthogonal noncoherent frequency-shift keying (NCFSK) with N branch signal-plus-noise (S + N) selection combining (SC) in Nakagami-m fading is studied. Both independent, identically distributed and independent, nonidentically distributed diversity branches are considered. Two S + N SC receiver structures are studied and their performances are analyzed. Tractable expressions are derived for the symbol error rate performance of these receiver structures in Nakagami-m fading with integer values of m. The performances of the S + N SC combiners are compared to those of classical SC and square-law combining (SLC). The effects of the modulation order, the fading parameter, the average decay factor and the number of diversity branches on the performance of S + N SC are examined and compared to the cases of classical SC and SLC. Monte Carlo simulation results are presented to test the validity of our analytical expressions.

In this paper, we consider the performance of M-ary orthogonal noncoherent frequency shift keying (NCFSK) with two signal-plus-noise (S + N) selection combining (SC) models in slow, flat Rician fading channels. We assume the fading branches are independent but not necessarily identically distributed. The probability of error for each model is derived as a single integral with finite integration limits. Extensive Monte Carlo simulation results are presented to assess the validity of our analytical expressions. Our results indicate that, unlike previously published results, the performance of S + N SC is not always superior to that of classical SC in fading and is dependent on the value of the signal-to-noise ratio, the modulation order, the multipath intensity profile and the fading parameter. The performances of S + N SC schemes are also compared to the performances of equal gain combining and square-law combining receivers.

Recent work has examined the performance of dual branch postdetection switch-and-stay combining (SSC) for noncoherent binary orthogonal frequency-shift keying (BFSK) and noncoherent M-ary orthogonal frequency-shift keying (MFSK) operating in the presence of slow, flat fading modeled by Rayleigh, Nakagami-m, and Rician distributions. These studies are important, in particular, because they quantify the improvement of postdetection SSC over predetection SSC. However, previous work on the performance of postdetection SSC over Rician fading is incorrect. In this paper, we present correct analyses of the bit error rate (BER) performance of BFSK and MFSK with dual-branch postdetection SSC in Rician fading. Extensive Monte Carlo simulations are carried out to verify the validity of the analytical results. Our results show that the performance gain of postdetection SSC over predetection SSC is much less than previously reported.

In two recent papers, the performances of dual-branch postdetection switch-and-stay combining (SSC) for noncoherent orthogonal binary frequency-shift keying (BFSK) and noncoherent orthogonal M-ary frequency-shift keying (MFSK) operating in the presence of slow flat fading modeled by Rayleigh, Nakagami-m, and Rician distributions have been analyzed. In this paper, we show that these previous analyses for the Nakagami-m fading model, which are restricted to integer values of m, are incorrect, and we derive the correct bit-error rate (BER) performances of BFSK and MFSK with dual-branch SSC in Nakagami-m fading for all values of m. Optimum switching thresholds that minimize the BER of BFSK and MFSK with postdetection SSC in Nakagami-m fading are obtained. The performance of postdetection SSC is compared with the performance of predetection SSC, and it is shown that postdetection SSC outperforms predetection SSC for all values of signal-to-noise ratio (SNR). We also show that for a given BER, the performance gain of postdetection SSC over predetection SSC has been overestimated by several decibels in SNR in previous publications.

In traditional selection combining, the branch having the largest signal-to-noise ratio is employed for signal detection. However, many practical systems choose the branch with the largest signal-plus-noise (S+N) at the filter output. Recent work has studied the performance of noncoherent orthogonal frequency shift keying (NCFSK) with S+N selection combining in slow, flat Rayleigh fading channels. We consider the performance of NCFSK with dual-branch S+N selection combining in slow, flat Rician and Nakagami-m fading channels. Two S+N receiver structures for NCFSK are examined and analytical expressions for the bit error rate (BER) performances of these receivers are obtained both for Rician and Nakagami-m fading channels. The channel fadings are assumed to be independent, but not necessarily identically distributed. It is shown that both the S+N receiver structures considered outperform the traditional selection combiner and their performances are compared to the performances of equal gain and square law combiners.

Lower and upper bound expressions for the penalty of hybrid diversity used with any two-dimensional (2-D) signaling constellation with polygonal decision regions are derived. Using these bounds with well-known results for maximal ratio combining, the error probability performance of 2-D signaling with hybrid selection/maximal ratio combining can be approximated to a high accuracy.

In hybrid selection/maximal ratio combining the receiver selects the L branches with the largest signal-to-noise ratios from N available diversity branches and performs maximal ratio from N available diversity branches and performs maximal ratio combining. A penalty is incurred with respect to maximal ratio combining which can be defined as the increase in signal-to-noise ratio required for hybrid combining to achieve the same symbol error probability as maximal ratio combining. In this paper, we derive the asymptotic signal-to-noise ratio penalties for small and large values of signal-to-noise ratio for arbitrary signaling constellations in Rayleigh fading. We also present new analytical expressions for the symbol error probabilities of 4-ary orthogonal signaling and 8-ary biorthogonal signaling with hybrid diversity combining in Rayleigh fading.

In hybrid selection/maximal ratio combining the receiver selects the L branches with the largest signal-to-noise ratios from N available diversity branches and performs maximal ratio combining. This incurs a loss or penalty with respect to maximal ratio combining which can be defined as the increase in signal-to-noise ratio required for hybrid combining to achieve the same symbol error probability as maximal ratio combining. In this paper, we derive lower and upper bound expressions for the penalty of any two-dimensional signaling constellation with polygonal decision regions. Using these bounds with well known results for maximal ratio combining, the error probability performance of two-dimensional signaling with hybrid diversity can be approximated at small and large values of signal-to-noise ratio, to a high accuracy.