The power consumption in commercial processors and application specific integrated circuits increases with decreasing technology nodes. Power saving techniques have become a first class design point for current and future VLSI systems. These systems employ large on-chip SRAM memories. Reducing memory leakage power while maintaining data integrity is a key criterion for modern day systems. Unfortunately, state of the art techniques like power-gating can only be applied to logic as these would destroy the contents of the memory if applied to a SRAM system. Fortunately, previous works have noted large temporal and spatial locality for data patterns in commerical processors as well as application specific ICs that work on images, audio and video data. This thesis presents a novel column based Energy Compression technique that saves SRAM power by selectively turning off cells based on a data pattern. This technique is applied to study the power savings in application specific inegrated circuit SRAM memories and can also be applied for commercial processors. The thesis also evaluates the effects of processing images before storage and data cluster patterns for optimizing power savings.

<p>The past decade has seen wide availability of solid-state drives (SSDs) in settings ranging from personal computing to enterprise storage. Their success over the hard disks is driven by performance considerations and cost savings. Besides SSDs based on flash memory, there have been ongoing efforts in developing other non-volatile memory technologies such as phase-change memory and MRAM. All these technologies enable what we refer to as random-access block devices. Unlike hard disks, these devices have fast random accesses; on the other hand, their writes are more expensive than their reads. In this work, we study how to optimize database and storage algorithms for the I/O characteristics of random-access block devices. Specifically, we tackle the following three problems.</p><p>The first one is about permuting data out-of-core. While external merge sort is popular for implementing permutation on hard disks, it carries unnecessary overhead in storing and comparing keys. We propose more efficient algorithms for a useful class of permutations called Address-Digit Permutations on random-access block devices.</p><p>The second problem is concurrency control for indexes on SSDs. Various indexes have been proposed for these devices, but to make such indexes practical, we must address the issue of concurrency control. We propose a novel indexing and concurrency control scheme which allows concurrent accesses during ongoing index reorganizations, and does so with minimal memory and block-level locking.</p><p>The third problem concerns log-structured merge, a popular indexing technique well-suited to random-access block devices. We show how an intelligent partial merge policy, combined with a block-preserving merge procedure, can ­significantly lower write traffic while preserving other advantages of log-structured merge.</p> / Dissertation

The gradually widening speed disparity of between CPU and memory has become an overwhelming bottleneck for the development of Chip Multiprocessor (CMP) systems. In addition, increasing penalties caused by frequent on-chip memory accesses have raised critical challenges in delivering high memory access performance with tight power and latency budgets. To overcome the daunting memory wall and energy wall issues, this thesis focuses on proposing a new heterogeneous scratchpad memory architecture which is configured from SRAM, MRAM, and Z-RAM. Based on this architecture, we propose two algorithms, a dynamic programming and a genetic algorithm, to perform data allocation to different memory units, therefore reducing memory access cost in terms of power consumption and latency. Extensive and intensive experiments are performed to show the merits of the heterogeneous scratchpad architecture over the traditional pure memory system and the effectiveness of the proposed algorithms.

The main objective of this research is to provide an efficient and reliable method for using multi-level cell (MLC) phase-change memory (PCM) as a main memory. As DRAM scaling approaches the physical limit, alternative memory technologies are being explored for future computing systems. Among them, PCM is the most mature with announced commercial products for NOR flash replacement. Its fast access latency and scalability have led researchers to investigate PCM as a feasible candidate for DRAM replacement. Moreover, the multi-level potential of PCM cells can enhance the scalability by increasing the number of bits stored in a cell. However, the two major challenges for adopting MLC PCM are the limited write endurance cycle and the resistance drift issue. To alleviate the negative impact of the limited write endurance cycle, this thesis first introduces a secure wear-leveling scheme called Security Refresh. In the study, this thesis argues that a PCM design not only has to consider normal wear-out under normal application behavior, most importantly, it must take the worst-case scenario into account with the presence of malicious exploits and a compromised OS to address the durability and security issues simultaneously. Security Refresh can avoid information leak by constantly migrating their physical locations inside the PCM, obfuscating the actual data placement from users and system software. In addition to the secure wear-leveling scheme, this thesis also proposes SAFER, a hardware-efficient multi-bit stuck-at-fault error recovery scheme which can function in conjunction with existing wear-leveling techniques. The limited write endurance leads to wear-out related permanent failures, and furthermore, technology scaling increases the variation in cell lifetime resulting in early failures of many cells. SAFER exploits the key attribute that a failed cell with a stuck-at value is still readable, making it possible to continue to use the failed cell to store data; thereby reducing the hardware overhead for error recovery. Another approach that this thesis proposes to address the lower write endurance is a hybrid phase-change memory architecture that can dynamically classify, detect, and isolate frequent writes from accessing the phase-change memory. This proposed architecture employs a small SRAM-based Isolation Cache with a detection mechanism based on a multi-dimensional Bloom filter and a binary classifier. The techniques are orthogonal to and can be combined with other wear-out management schemes to obtain a synergistic result. Lastly, this thesis quantitatively studies the current art for MLC PCM in dealing with the resistance drift problem and shows that the previous techniques such as scrubbing or error correction schemes are incapable of providing sufficient level of reliability. Then, this thesis proposes tri-level-cell (3LC) PCM and demonstrates that 3LC PCM can be a viable solution to achieve the soft error rate of DRAM and the performance of single-level-cell PCM.

In this thesis, CMs of linear and non-linear multiple antenna receivers, in particular linear minimum mean squared error (LMMSE) and LMMSE with decision feedback (LMMSE-DF), are developed. To develop these CMs, first a simple analytical expression of the distribution of the post processing signal to interference and noise (SINR) of an LMMSE receiver is developed. This expression is then used to develop SINR- and ABER-based CMs. However, the analytical forms of these CMs are derived only for the following scenarios: (i) any number of receive antennas with three users having arbitrary received powers and (ii) two antenna receiver with arbitrary number of equal received power users. For all the other scenarios a semi-analytical CM is used. The PHY abstractions or CMs are next used in the evaluation of a random access cellular network and an ad hoc network. Analytical model of the random access cellular network is developed using the SINR- and ABER-based CM of the LMMSE receiver. The impact of receiver processing is measured in terms of throughput. In this case, the random access mechanism is modeled by a single channel S-Aloha channel access scheme. Another analytical model is developed for single and multi-packet reception in a multi-channel S-Aloha channel access. An emph{ideal} receiver is modeled in this case, i.e. the packet(s) are successfully received as long as the total number of colliding packets is not greater than the number of antennas. Throughput and delay are used as performance metrics to study the impact of different PHY designs. Finally, the SINR-based semi-analytical CMs of LMMSE and LMMSE-DF are used to evaluate the performance of multi-hop ad hoc networks. Throughput is used as the performance evaluation metric. A novel MAC, called S-MAC, is proposed and its performance is compared against another MAC for wireless networks, called CSMA/CA(k).

Aquesta tesi suggereix usar un compte enrere determinista després de les transmissions exitoses en les xarxes d'àrea local sense fils. Com que les estacions que han transmès amb èxit en el seu darrer intent no poden col·lisionar entre elles en el seu proper intent, el nombre de col·lisions es redueix. Per tant, anomenem el protocol accés múltiple per detecció de portadora amb evitament de col·lisions millorat. El protocol es modela i es simula en diversos escenaris per a la seva avaluació. S'arriba a la conclusió de que el protocol proposat ofereix un rendiment igual o major que l'existent. / This thesis suggests the use of a deterministic backoff after successful transmissions in the MAC protocol of WLANs. Since those stations that successfully transmitted in their last collision attempt cannot collide among them in their next transmission attempt, the number of collisions is reduced. Hence, the protocol is called Carrier Sense Multiple Access with Enhanced Collision Avoidance. The protocol is modelled and simulations are used to assess its performance in a variety of scenarios. It is concluded that the proposed protocol performs always equal or better than the legacy one.

Thesis (M.C.S.) - Carleton University, 2007. / Includes bibliographical references (p. 77-84). Also available in electronic format on the Internet.

Avec l'introduction des plate-formes CubeSat, le nombre de petits satellites lancés dans l'espace a grandi de manière importante pendant les deux dernières décennies.Étant développés initialement par des universités et des centres de recherche pour des simples tests technologiques ou des expériences académiques, ces plate-formes aujourd'hui permettent d'envisager de nouvelles applications et services.Dans cette thèse, nous nous intéressons à l'usage de petits satellites à défilement pour des réseaux globaux de collecte de données et, plus généralement, pour des applications de type machine-to-machine (M2M).En raison des contraintes existantes tant au segment sol comme au segment spatial, la capacité du canal de transmission est fortement limitée---notamment celle du lien montant, qui correspond à un canal à accès multiple.Ces réseaux sont aussi caractérisés par des très petits messages arrivant au système de manière imprévisible, ce qui implique que toute redondance liée au protocole a un impact important sur l’efficacité spectrale. Ainsi, des méthodes d'accès aléatoires sont souvent préférés pour le lien montant.Relever ces défis nécessite d'aborder l'optimisation de la transmission de manière holistique. Plus spécifiquement, la conception des couches physiques (PHY) et de contrôle d'accès au support (MAC, de l'anglais Media Access Control) doit être menée de manière conjointe.Les principales contributions de cette thèse portent sur l'étude du protocole Time-- and Frequency--Asynchronous ALOHA (TFAA), une technique d'accès aléatoire utilisée dans des réseaux terrestres à modulation de bande étroite. En réduisant significativement le débit binaire de transmission, TFAA permet notamment d'établir des liaisons à longue portée et/ou à faible consommation énergétique, dont des systèmes M2M par satellite sont un exemple.D'abord, nous évaluons les performances au niveau MAC (i.e., le taux d'utilisation de canal et la probabilité d'erreur de packet) sous trois différents modèles de réception: le modèle de collisions, le modèle de capture et un modèle plus détaillé qui prend en compte les paramètres de la couche PHY.À partir de ce dernier modèle, nous étudions ensuite l'impact de certains paramètres de la couche PHY sur les performances au niveau MAC.Afin d'améliorer la performance de TFAA, nous proposons Contention Resolution Time-- and Frequency--Asynchronous ALOHA (CR-TFAA), une solution plus sophistiquée intégrant des techniques de suppressions successives d'interférences.Enfin, nous étudions les bénéfices obtenus en exploitant le compromis <<performance--délai de bout-en-bout>> en utilisant des techniques simples telles qu'un système de contrôle de transmission et le codage au niveau packet. / With the introduction of the CubeSat standard, the number of small-satellite missions has increased dramatically over the last two decades.Initially developed by universities and research centres for technology validation and academic experiments, these low-cost platforms currently allow to perform a variety of advanced, novel applications.In this thesis we are interested in the use of small satellites for global data collection and, more generally, for Internet of Things (IoT) and machine-to-machine (M2M) applications.Since both the space and ground segments are subject to stringent constraints in terms of size and mass, the overall capacity of the communications channel is highly limited, specially that of the uplink, which is a multi-access channel.These systems are also characterised by bursty, short messages, meaning that any protocol overhead may have a significant impact on the bandwidth efficiency. Hence, a random access approach is usually adopted for the uplink.Facing these challenges requires to optimize the communication system by taking an holistic approach. In particular, a joint design of both the physical (PHY) and Medium Access Control (MAC) layers is needed.The main contributions of this thesis are related to the study of Time-- and Frequency--Asynchronous ALOHA (TFAA), a random access approach adopted in terrestrial ultra narrowband (UNB) networks. By trading data rate for communication range or transmission power, TFAA is particularly attractive in power constrained applications such as low power wide area networks and M2M over satellite. First, we evaluate its MAC performance (i.e., its throughput and packet error rate) under three different reception models: the collision channel, the capture channel and a more detailed model that takes into account the PHY layer design.Then, we study the impact of PHY layer parameters, such as forward error correction (FEC), pulse shaping filter and modulation order, on the MAC performance.We show that, due to the characteristics of the multiple access interference, significant improvements can be obtained by applying low-rate FEC.To further improve TFAA's performance, we propose Contention Resolution Time-- and Frequency--Asynchronous ALOHA (CR-TFAA), a more advanced design which is in line with recent developments such as Asynchronous Contention Resolution Diversity ALOHA (ACRDA).Under the same set of hypothesis, we see that CR-TFAA provides similar and even better performance than ACRDA, with a decrease in the packet error rate of at least one order of magnitude.Finally, we study the benefits that can be obtained by trading delay for MAC performance and energy efficiency, using simple techniques such as transmission control and packet-layer erasure coding.

京都大学 / 0048 / 新制・論文博士 / 博士(工学) / 乙第13015号 / 論工博第4140号 / 新制||工||1650(附属図書館) / 32943 / (主査)教授 斧 髙一, 教授 木村 健二, 教授 立花 明知 / 学位規則第4条第2項該当 / Doctor of Philosophy (Engineering) / Kyoto University / DFAM

Massive MIMO communication systems have been highlighted as the main technology for physical layer of next generation communication standards, like 5G. While conventional communication between BS and its covered users is performed in orthogonal time-frequency resources, the improved interuser interference mitigation capability provided by the large number of BS antennas enables the BS to communicate with several users in the same time-frequency resource. This better usage of available but scarce spectrum elevates the spectral efficiency to very appreciable levels, and has a similar effect on energy efficiency, since the transmit power is not increased. On the other hand, if the objective is to provide a target performance for the users, the required transmit power in both direct and reverse links can be made inversely proportional to the number of BS antennas employed. In this Doctoral Thesis, several important aspects of massive MIMO systems are systematically investigated aiming to improve their energy and spectral efficiencies. We can enumerate our main contributions as follows. Considering a cellular massive MIMO network, we proposed an optimized assignment policy of training sequences to the users, which is then combined with suitable power control algorithms. We have also investigated the adoption of alternative waveforms in this scenario, such as single-carrier transmission, in order to overcome the issues of conventional OFDM. Our contributions in this topic are to derive analytical performance expressions for a time-domain single-carrier equalizer taking advantage of the large number of BS antennas, and to evaluate and compare the total energy efficiency of OFDM versus single-carrier massive MIMO systems. Finally, considering crowded massive MIMO networks, composed by both human users as well as machine-type communication devices, we proposed an improved random access protocol aiming to decrease the average number of access attempts for the users and decreasing the probability of failed access attempts. / Sistemas de comunicação de múltiplas antenas (multiple-input multiple-output - MIMO) têm se destacado como a principal tecnologia para a camada física dos padrões de comunicação da próxima geração, como o 5G. Enquanto a comunicação convencional entre a estação base (base station - BS) e seus usuários atendidos é realizada em recursos ortogonais de tempo-frequência, a grande capacidade de redução da interferência interusuários possibilitada pelo grande número de antenas da BS habilita a BS a se comunicar com diversos usuários no mesmo recurso tempo-frequência. Este melhor uso do escasso espectro disponível eleva a eficiência espectral a níveis muito apreciáveis, e tem um efeito similar na eficiência energética, pois a potência de transmissão não é aumentada. Por outro lado, se o objetivo é fornecer um desempenho desejado para os usuários, a potência de transmissão necessária em ambos os enlaces direto e reverso pode ser feita inversamente proporcional ao número de antenas na BS. Nesta Tese de Doutorado, diversos aspectos importantes de sistemas MIMO massivo são sistematicamente investigados com o objetivo de melhorar suas eficiências energética e espectral. Pode-se enumerar as principais contribuições alcançadas como se segue. Considerando uma rede celular MIMO massivo, propõe-se uma política de atribuição de sequências de treinamento aos usuários otimizada, a qual é depois combinada com apropriados algoritmos de controle de potência. Também investiga-se a adoção neste cenário de formas de onda alternativas, tal como a transmissão de portadora única, visando superar as deficiências da convencional multiplexagem por divisão de portadoras ortogonais (orthogonal frequency-division multiplexing - OFDM). As principais contribuições obtidas neste tema são derivar expressões de desempenho analíticas para um equalizador de portadora única no domínio do tempo que aproveita o grande número de antenas na BS, e avaliar e comparar a eficiência energética total de sistemas MIMO massivo OFDM versus portadora única. Finalmente, considerando redes MIMO massivo sobrecarregadas, compostas por usuários humanos bem como dispositivos de comunicação do tipo máquina, propõe-se um protocolo de acesso aleatório melhorado visando diminuir o número médio de tentativas de acesso para os usuários e diminuir a probabilidade de falhas de tentativa de acesso.