We investigate packet discarding schemes for TCP over ATM with GFR service. We test the throughput, fairness and goodput of two existing schemes, EPD and DFBA. In the simulation, we find that goodput is low when the network is incongestion. To provide a solution for this problem, we propose the SWFA(Selective Weighted Fair Allocation), which discards packets from selected sessions rather than by probability. Our results shows that SWFA achieves throughput and goodput enhancement over other schemes.

Nowadays, X86 CPU all have superscalar computing ability. Superscalar architecture can fetch, execute and commit more than one instruction per cycle. And it helps a lot to explore more instruction level parallelism. If a superscalar processor fetches instructions inefficiently, its performance speedup ratio will be limit. Program flow is not continuous. It is one of main reasons that Front-End can¡¦t fetch efficiently. And it is useless to get more speedup by enlarging fetch capacity of Front-End or other units. In this thesis, we present a new structure of branch target buffer and instruction stream buffer. They have abilities to predict advance branch information and reassemble cache lines. Front-End could fetch more valid instructions in a cycle by reassembling original line and line which contains instructions of the next basic block. The simulation and implement results show that we can get 43.2% speedup in fetch efficiency with 64 bytes cache line size and 6 fetch capacities. And 3.6 valid instructions per cycle with ABP buffer which buffers 4 cache line.

Following the surface application of different fertiliser treatments (cattle slurry, cattle manure, and inorganic fertiliser) to grassed hillslopes, the transport of nitrogen (N) and phosphorus (P) was monitored on controlled plots. The plots measured 30 m by 5 m, the lower 10 m acting as an untreated buffer zone, in order to examine the effect of following current codes of practice when applying fertilisers (MAFF and WOAD, 1991). Surface and sub-surface nutrient flow pathways were monitored, by sampling water from the saturated and unsaturated zones of the soil profile, and from surface runoff. Sub-surface flow was dominated by N03--N. However, concentrations of this N species were relatively low (3-6 mg 1-1), and differences between treatments and the control were not significant. Generally, this indicated immobilisation of N species and P fractions, outputs in plant uptake and possibly to atmosphere (denitrification). For surface runoff, concentrations of N and P from treated plots were significantly higher than those from the control. N was largely present as organic-N and NH4+-N for the slurry and manure (15 mg 1-1 and 5 mg 1-1 respectively), and as NH4+-N and N03--N for the inorganic fertiliser (20 mg 1-1 in both cases). P was largely present as P04--P (0.5 mg 1-1 for manure and slurry, 10 mg 1-1 for inorganic fertiliser), except for the manure treatment, where some 75 % of the total was organic-P. The 10 m buffer was effective in reducing the delivery of N and P in surface runoff, differences between the treatments and the control being rendered insignificant. This result was qualified in terms of the ratio of the buffer area to the treated area and the relatively low surface loads of N and P observed. N and P transport was then examined under less controlled conditions at the field to headwater catchment scale (4-18 ha). Sub-surface throughflow was associated with similar concentrations of N03--N (3-8 mg 1-1) and very low concentrations of P04--P (0.03-0.1 mg 1-1). These concentrations appeared to vary more in relation to seasonal changes in the level of discharge from the catchment than to changes in land-use, which implied that N and P in sub-surface throughflow was transport rather than supply limited. This was in broad agreement with observations made at the plot scale. At this larger scale, fluctuations in the concentrations of NH4+-N and P04--P occurred during rainfall events in response to what was probably the short-term occurrence of surface runoff from partial source areas of reduced infiltration capacity or variable source areas of saturation. These short-term fluctuations implied variations in topography and the distribution of surface derived N and P at the catchment scale. There was evidence of the occurrence of preferential flow during field experiments at both scales, which had important implications for the timing and magnitude of N and P transport. Finally, a functional and semi-distributed mathematical model was constructed to operate at the headwater catchment scale, and the effect of spatial variability in the interaction between land-use and topography on the transport of N and P was considered further.

Made available in DSpace on 2019-03-29T23:53:05Z (GMT). No. of bitstreams: 0 Previous issue date: 2015-10-05 / For decades, hard disk drives (HDD) have dominated the scene with regard to data storage for large databases systems. HDDs may be considered a symmetric media because there is no difference between the time (and cost) for reading and writing data. As a counterpart, in the last years, a whole new class of storage media has raised, whose the main feature is to have no mechanical parts and, more importantly, to be asymmetric. In a asymmetric media, the time forreading data is faster than the time in which such media writes data. Depending on manufacturing, this asymmetry may reach a factor of 10 or even higher. Asymmetric storage impacts on the most important database management system components, more specifically, on the buffer manager. In this research, we propose database buffer replacement algorithms for asymmetric media. They try to take advantage from the use of a asymmetric media by keeping written(dirty) pages in main memory, postponing their writing down on media and also adapting to achieve good hit ratios. Keywords: Databases, Storage Class Memory, Buffer Management Policies. / Durante décadas, unidades de disco rígido (HDD) têm dominado o armazenamento de dados para grandes sistemas gerenciadores de banco de dados. Os HDDs podem ser considerados um meio simétrico porque não há diferença entre o tempo (custo) das operações de leitura e escrita. Em contrapartida, nos últimos anos, uma nova categoria de dispositivos de armazenamento ganhou notoriedade. Sua principal característica é não possuir partes mecânicas e, mais importante, serem assimétricos. Em uma mídia assimétrica, o tempo para a leitura de dados é bastante inferior ao tempo gasto por uma operação de escrita. A depender do fornecedor, a assimetria pode atingir um fator de 10 ou mais. Essa assimetria gera um impacto direto nos componentes mais importantes do SGBD, mais especificamente, no gerenciador de buffer. Neste trabalho, propomos algoritmos de substituição de páginas para mídias assimétricas. Os algoritmostentam se beneficiar da mídia assimétrica mantendo as páginas alteradas em memória principal, adiando a sua escrita para a mídia de armazenamento persistente e adaptando-se para também obter boas taxas de acerto. Palavras-chave: Bancos de Dados, Mídias Assimétricas, Políticas de Gerenciamento de Buffer.

In this thesis, we propose two mechanisms, stream splitting and merging, to fully utilize the bandwidth in a mixed wired and wireless network. The two mechanisms implemented on a SVC Streaming Gateway (SSG) mainly exploit the characteristics of multi-layer SVC (Scalable Video Coding). In the stream splitting mechanism, the available bandwidth can be calculated on SSG, so that an adequate number of layers of a SVC stream can be determined for splitting. In the stream merging mechanism, when the available bandwidth is increased, a stream with more number of enhancement layers can be extracted to merge onto a stream with less number of layers. Additionally, when two streams pass through SSG concurrently, one stream may arrive at SSG faster due to different traffic conditions. If one stream with more SVC layers is faster than the one with less SVC layers, we buffer the packets of the faster stream on SSG, so that the buffered packets can be merged to the top of SVC layers of the slow stream. On the other hand, if one stream with more SVC layers is slower than the one with less SVC layers, to effectively merge the enhancement layers of the slow stream onto the faster stream, SSG will delay the packets of the faster stream to wait for the packets of the slow stream. Finally, we implement these two mechanisms on the Linux platform. In the experiments, we measure buffer delay, queue length, and the number of merged packets to demonstrate that the two proposed mechanisms can effectively improve the picture freeze-up and mosaic phenominon.

According to the literature, the choice of the material, the device structure and the morphology of the active layer play an important role in the performance of the organic solar cell. For instance, many groups insert buffer layer between cathode and active layer to improve device efficiency. Otherwise, the tandem structure have attracted much attention due to different absorbing spectra of the top and bottom cells which can improve the device efficiency. In this study, we select TiOPc as the interlayer material because it can both extend the device absorption spectra and improve device performance. In our results, the thickness of 1 nm TiOPc has the best device efficiency based on the device structure ITO(1750 Å)/PEDOT:PSS(500 Å)/ P3HT:PCBM(800 Å)/TiOPc(10 Å)/Al(2000 Å) and the device area of 0.16 cm2, the performance increase from 0.97 % to 1.77 % compared to standard device under AM1.5 simulation light before post annealing. (which is 82.4 % improved). This consequence can be attributed to a increase open voltage from 0.40 V to 0.56 V, a short increas current density from 5.62 mA/cm2 to 6.16 mA/cm2 and fill factor increased from 0.428 to 0.511. After post annealing at 140 ¢XC, the power efficiency increase for 45.2 % contrast to standard post annealing.

The materials based on GaN have successfully developed on short-wavelength laser diodes (LDs), light-emitting diodes (LEDs) and ultraviolet photodetector. In this study, GaN epitaxial layers have been successfully grown on sapphire substrates. We used several methods including the nitridation temperature for substrates before growing epilayer, the growth temperature and time of buffer layer and the growth temperature of GaN epilayer to study it. From the results of the photoluminescence (PL) measured at 77K, the X-Ray diffraction measurement, SEM cross sectional views to realize the characteristic and we get a better qualities of GaN epilayers after using the foregoing methods. In this study, the re-crystallization of the buffer layer would occur while temperature re-rise to high temperature, and the phenomenon of conglomeration influence the quality and morphology of GaN epilayers. According to the results of the experiments, we study the mechanisms of yellows luminescence and donor-acceptor pair.

Buffer insertion is an essential technique for reducing interconnect delay in submicron circuits. Though it is a well researched area, there is a need for robust and effective algorithms to perform buffer insertion at the circuit level. This thesis proposes a new buffer insertion algorithm for large circuits. The algorithm finds a buffering solution for the entire circuit such that buffer cost is minimized and the timing requirements of the circuit are satisfied. The algorithm iteratively inserts buffers in the circuit improving the circuit delay step by step. At the core of this algorithm are very simple but extremely effective techniques that constructively guide the search for a good buffering solution. A flexibility to adapt to the user's requirements and the ability to reduce the number of buffers are the strengths of this algorithm. Experimental results on ISCAS85 benchmark circuits show that the proposed algorithm, on average, yields 36% reduction in the number of buffers, and runs three times faster than one of the best known previously researched algorithms.

published_or_final_version / Political Science / Master / Master of Philosophy

No description available.