771 research outputs found
Data Structures for Modern Memory and Storage Hierarchies (Dagstuhl Seminar 21283)
No full textThis report documents the program and the outcomes of Dagstuhl Seminar 21283 "Data Structures for Modern Memory and Storage Hierarchies". For decades, computers consisted of a CPU, volatile main memory, and persistent disk. Today, modern storage technologies such as flash and persistent memory as well as the seemingly inevitable migration into virtualized cloud instances, connected through high-speed networks, have radically changed the hardware landscape. These technologies have major implications on how to design data structures and high-performance systems software. The seminar discussed how to adapt data structures and software systems to this new hardware landscape
Study of emission of secondary particles during ion scuttering and its effects on LEIS spectra
No full textV tejto práci skúmame emisiu sekundárnych častíc pomocou spektroskopie rozptylu iónov s nízkou energiou (LEIS) a rozptylu iónov so strednou energiou (MEIS). LEIS využíva primárny zväzok iónov vzácnych plynov s počiatočnými energiami niekoľkých keV a je obzvlášť citlivý na najvzdialenejšiu povrchovú vrstvu. Skúmané sú zmeny povrchovej kontaminácie medi, platiny a kremíka, pričom hodnotíme vplyv čistenia rozprašovaním a žíhania na prirodzene sa vyskytujúcu aj indukovanú kontamináciu plynom CO. Tiež sledujeme vplyv pasivácie vodíkom na povrch kremíku. Okrem toho na štúdium desorpcie povrchovej kontaminácie z monokryštalickej Si(100) samonosnej membrány s hrúbkou 50 nm použijeme metódu ToF MEIS S energetickým rozsahom do niekoľkých stoviek keV. Primárne iónové zväzky He, Ne a B sú použité na skúmanie membrány v štandardnej geometrii spätného rozptylu, ale aj v novej priechodnej geometrii experimentu. To nám umožňuje analyzovať emisiu sekundárnych častíc s oboch povrchov. Desorpcia povrchového znečistenia sa predtým pripisovala výhradne energii predanej elektrónovému systému. Ťažšie ióny s vyššou hodnotou straty energie predanej medzi jadrami, však zvyšujú elektronickú depozíciu energie, čo odhaľuje ich synergický efekt. Kombinované účinky zvýšeného prenosu energie do jadrového podsystému a smerový charakter kaskády zrážok zvyšujú výťažok sekundárnych iónov. Tieto poznatky môžu výrazne posunúť výskum čistenia a štruktúrovania 2D materiálov pomocou iónových zväzkov, ktoré ďalej umožní modifikáciu povrchu a kontrolu kontaminácie.In this thesis, we investigate the emission of secondary particles using both Low Energy Ion Scattering (LEIS) and Medium Energy Ion Scattering (MEIS) spectroscopy. LEIS utilizes a primary beam of noble gas ions with initial energies of a few keV and is particularly sensitive to the outermost surface layer. We examine the surface contamination changes on copper, platinum, and silicon, assessing the effects of sputter cleaning and annealing on both naturally occurring and induced CO gas contamination. Additionally, we employ Time-of-Flight MEIS with energy ranges up to a few hundred keV to study the desorption of surface contamination from a single-crystalline Si(100) self-supporting 50 nm thick membrane. Primary beams of He, Ne, and B were used to probe the membrane in both standard backscattering and novel transmission measurement geometries, allowing us to analyze emission from both surfaces. The desorption of surface contamination was previously attributed to electronic sputtering. However, heavier ions with higher nuclear stopping enhance the electronic energy deposition, revealing a synergy effect between them. The combined effects of increased energy deposition to the nuclear subsystem and the directional nature of the collision cascade increased the yield of secondary ions. This knowledge can significantly advance the precision cleaning and structuring of 2D materials using ion beams, enabling surface modification and control of contamination.
Study of emission of secondary particles during ion scuttering and its effects on LEIS spectra
No full textV tejto práci skúmame emisiu sekundárnych častíc pomocou spektroskopie rozptylu iónov s nízkou energiou (LEIS) a rozptylu iónov so strednou energiou (MEIS). LEIS využíva primárny zväzok iónov vzácnych plynov s počiatočnými energiami niekoľkých keV a je obzvlášť citlivý na najvzdialenejšiu povrchovú vrstvu. Skúmané sú zmeny povrchovej kontaminácie medi, platiny a kremíka, pričom hodnotíme vplyv čistenia rozprašovaním a žíhania na prirodzene sa vyskytujúcu aj indukovanú kontamináciu plynom CO. Tiež sledujeme vplyv pasivácie vodíkom na povrch kremíku. Okrem toho na štúdium desorpcie povrchovej kontaminácie z monokryštalickej Si(100) samonosnej membrány s hrúbkou 50 nm použijeme metódu ToF MEIS S energetickým rozsahom do niekoľkých stoviek keV. Primárne iónové zväzky He, Ne a B sú použité na skúmanie membrány v štandardnej geometrii spätného rozptylu, ale aj v novej priechodnej geometrii experimentu. To nám umožňuje analyzovať emisiu sekundárnych častíc s oboch povrchov. Desorpcia povrchového znečistenia sa predtým pripisovala výhradne energii predanej elektrónovému systému. Ťažšie ióny s vyššou hodnotou straty energie predanej medzi jadrami, však zvyšujú elektronickú depozíciu energie, čo odhaľuje ich synergický efekt. Kombinované účinky zvýšeného prenosu energie do jadrového podsystému a smerový charakter kaskády zrážok zvyšujú výťažok sekundárnych iónov. Tieto poznatky môžu výrazne posunúť výskum čistenia a štruktúrovania 2D materiálov pomocou iónových zväzkov, ktoré ďalej umožní modifikáciu povrchu a kontrolu kontaminácie
Concurrency Control for High-Performance Storage Engines
No full textHistorically, storage engines were constrained by limited DRAM capacity and the slow hard disk drives, which led to the development of disk-based system architectures.
The emergence of many-core CPUs, along with the increased affordability and performance of PCIe-attached NVMe SSDs, has led to the development of high-performance storage engines.
These engines demonstrate that with a redesigned buffer manager, systems can achieve in-memory performance levels when the working set fits within the buffer pool, while gracefully degrading with larger data sets.
However, to fully exploit modern hardware, a complete redesign of all components of traditional storage engines is necessary.
In this thesis, we rethink concurrency control for this new emerging class of high-performance storage engines.
At a high level, we focus on Multi-Version Concurrency Control (MVCC), specifically snapshot isolation commit protocol, dependency tracking, version storage, and garbage collection.
On a lower level, we address the scalability and space utilization of B-Tree indexes in storage engines.
For scalable MVCC, we propose OSIC, a commit protocol for snapshot isolation that efficiently scales to more concurrent transactions than the number of CPU cores by minimizing cross-worker communication.
This allows us to harness the increasing number of IOPS that modern flash drives offer.
Additionally, we suggest a commit dependency tracking mechanism that combines distributed logging with Early Lock Release to reduce abort rate and latency.
For robust performance in mixed workloads, we propose a storage layer that automatically adjusts to user workloads at the tuple level without requiring manual user intervention.
This includes removing tombstones from the OLTP hot path and an adaptive version storage and garbage collection scheme.
We also address B-Tree node unnecessary write contention through node splitting and improve space utilization by merging underfull sibling nodes at eviction.Historisch gesehen waren Speichersysteme (Storage Engines) durch begrenzte DRAM-Kapazität und die langsame Festplattenlaufwerke eingeschränkt, was zur Entwicklung von diskbasierten Systemarchitekturen führte.
Das Aufkommen von Many-Core-CPUs sowie die gestiegene Erschwinglichkeit und Leistung von PCIe-gebundenen NVMe-SSDs führten zur Entwicklung von Hochleistungsspeichersystemen.
Diese Engines zeigen, dass Systeme mit einem neu gestalteten Puffermanager nahezu in-Memory-Performance erreichen können, wenn der Arbeitsdatensatz in den Puffer passt, und sich bei größeren Datensätzen mäßig verlangsamen.
Um jedoch die moderne Hardware voll auszuschöpfen, ist eine vollständige Neugestaltung aller Komponenten herkömmlicher Speichersysteme erforderlich.
In dieser Arbeit gestalten wir die Nebenläufigkeitskontrolle für diese aufstrebende Klasse von Speichersystemen neu.
Auf einer hohen Ebene konzentrieren wir uns auf die Multi-Version Concurrency Control (MVCC), insbesondere auf das Snapshot Isolation Commit Protokoll, die Speicherung von Versionen, die Abhängigkeitsverfolgung und Garbage Collection.
Auf einer niedrigeren Ebene adressieren wir die Skalierbarkeit und den Speicherplatzverbrauch von B-Baum Indexstrukturen in Speichersystemen.
Für skalierbare MVCC schlagen wir OSIC vor, ein Commit-Protokoll für Snapshot Isolation, das effizient skaliert und mehr nebenläufige Transaktionen als die Anzahl der CPU-Kerne ermöglicht, indem es die Kommunikation zwischen Workers minimiert.
Dies ermöglicht es uns, die steigende Zahl von IOPS zu nutzen, die moderne Flash-Laufwerke bieten.
Darüber hinaus schlagen wir einen Mechanismus zur Verfolgung von Commit-Abhängigkeiten vor, der verteiltes Logging mit Early Lock Release integriert, um die Abbruchrate und die Latenz der Transkationen zu reduzieren.
Für robuste Leistung in gemischten HTAP Workloads schlagen wir eine Storage Layer vor, die sich automatisch auf Benutzer-Workloads auf Tupel-Ebene anpasst, ohne manuelles Eingreifen des Benutzers zu erfordern.
Dies umfasst das Entfernen von Tombstones aus dem OLTP-Hot Path und ein adaptives Versions Storage und Garbage Collection Schema.
Wir adressieren auch unnötigen Schreibkonflikt auf B-Baum Knoten durch Knotenaufteilung und verbessern die Speichernutzung, indem wir unterausgelastete Geschwisterknoten beim Auslagern zusammenführen
MxTasks: a novel processing model to support data processing on modern hardware
No full textThe hardware landscape has changed rapidly in recent years. Modern hardware in today's servers is characterized by many CPU cores, multiple sockets, and vast amounts of main memory structured in NUMA hierarchies.
In order to benefit from these highly parallel systems, the software has to adapt and actively engage with newly available features.
However, the processing models forming the foundation for many performance-oriented applications have remained essentially unchanged.
Threads, which serve as the central processing abstractions, can be considered a "black box" that hardly allows any transparency between the application and the system underneath.
On the one hand, applications are aware of the knowledge that could assist the system in optimizing the execution, such as accessed data objects and access patterns.
On the other hand, the limited opportunities for information exchange cause operating systems to make assumptions about the applications' intentions to optimize their execution, e.g., for local data access.
Applications, on the contrary, implement optimizations tailored to specific situations, such as sophisticated synchronization mechanisms and hardware-conscious data structures.
This work presents MxTasking, a task-based runtime environment that assists the design of data structures and applications for contemporary hardware.
MxTasking rethinks the interfaces between performance-oriented applications and the execution substrate, streamlining the information exchange between both layers.
By breaking patterns of processing models designed with past generations of hardware in mind, MxTasking creates novel opportunities to manage resources in a hardware- and application-conscious way.
Accordingly, we question the granularity of "conventional" threads and show that fine-granular MxTasks are a viable abstraction unit for characterizing and optimizing the execution in a general way.
Using various demonstrators in the context of database management systems, we illustrate the practical benefits and explore how challenges like memory access latencies and error-prone synchronization of concurrency can be addressed straightforwardly and effectively
O Conceito de organismo no pensamento Kantiano
No full textDissertação (mestrado) - Universidade Federal de Santa Catarina, Centro de Filosofia e Ciências Humanas, Programa de Pós-Graduação em Filosofia, Florianópolis, 2010A biologia, assim como as demais ciências, possui um objeto de estudo, o organismo, que exige conceitos e métodos distintos para sua investigação. Muitas vezes, o avanço de uma ciência se deve ao fato de, diante do mesmo objeto de estudo, os investigadores servirem-se de novos conceitos que possibilitam a proposição de novas perguntas, a partir das quais é possível avançar no conhecimento. Na história da biologia, o conceito kantiano de organismo representou uma contribuição que possibilitou a proposição de novas perguntas, que permitiram novos avanços. O presente estudo almeja apresentar tal conceito kantiano como um elemento fecundo que emerge de sua crítica à Faculdade do Juízo. Ao longo da história da biologia, o conceito de organismo variou. Desde os antigos até os modernos, foi entendido de muitos modos, dependendo da concepção de mundo na qual estava inserido. Aristóteles, Spinoza, Descartes conceberam um conceito de organismo ajustado a sua metafísica. Tais conceitos foram criticados por Kant, por assentarem em uma metafísica dogmática. Do mesmo modo como seus predecessores, Kant concebeu um conceito de organismo ajustado a sua concepção, que, diferente das anteriores, de caráter dogmático, está marcada por uma abordagem crítico-transcendental, que encontra na relação do conhecimento com a faculdade cognitiva que o produz, a chave para compreender o "conhecer" e o "conhecimento" do objeto da biologia. Por isso, o conhecimento biológico, no pensamento Kantiano, encontra seu fundamento mais profundo e seguro no próprio sujeito cognoscente, no conceito de organismo, ou seja, de uma conformidade a fins objetiva material interna (fim natural), contributo do sujeito na "composição" do conhecimento biológico, que consiste numa regra heurística que orienta o refletir diante do múltiplo dado, sem a qual objetos tais como "seres organizados" não poderiam ser ajuizados. O conceito de organismo é uma modalidade do que Kant denominou "técnica da natureza", que faz pensar tanto a natureza, como objetos da natureza, como se fossem arte. Esse conceito analógico proporciona a regra da qual depende a Faculdade do Juízo Reflexiva para poder cumprir sua tarefa, qual seja, subsumir o dado múltiplo, elevando-se do particular ao universal. Tal faculdade, ao submeter o dado múltiplo ao conceito de "técnica da natureza", faz pensar a natureza determinada como sistema segundo leis empíricas e os objetos determinados como sistemas finais. O organismo, enquanto objeto da natureza, não é uma simples máquina, um agregado de partes, mas um sistema, cujo traço específico é ser causa e efeito de si mesmo (fim natural), capaz de autoproduzir-se enquanto substância, órgão, indivíduo e espécie. Por fim, realizaram-se algumas considerações sobre o impacto do conceito kantiano de organismo à concepção dos biólogos mecanicistas dos séculos XVII e XVIII, que equiparou o organismo à máquina. Segundo Kant, essa concepção faz pensar o organismo como expressão de uma teleologia artificial, por meio da qual a causalidade segundo fins é pensada como resultado da ação de um autor externo à obra, conduzindo a faculdade de conhecer a uma causa transcendente, ilegítima em uma abordagem crítico-transcendental do conhecimento. Contudo, defende Kant, o organismo é expressão de uma teleologia natural, por meio da qual a causalidade segundo fins é pensada meramente por analogia com a Arte, como se fosse efeito de uma causalidade conforme a fim. Assim, o conceito kantiano de organismo envolve uma teleologia que figura apenas como princípio heurístico à faculdade do conhecimento, sem o qual ela não seria capaz de conhecer as peculiaridades do organismo.Biology, like other sciences, has an object of study, the organism, which requires different concepts and methods for its investigation. So often, the advancement of science is due to the fact that, before the same privileged object of study, researchers make use of new concepts allowing the proposal of new questions, from which it is possible to advance in knowledge. In the history of biology, the Kantian concept of organism represented a contribution that enabled the proposal of new questions, that allowed new advances. This study aims at presenting such Kantian concept as a fruitful point that emerges from his critique of the faculty power of judgment. Throughout the history of biology, the concept of organism has varied. From the ancient to the modern, it has been understood in many ways, depending on the worldview within which it was inserted. Aristotle, Spinoza, Descartes devised a concept of organism adjusted to its metaphysics. Such concepts were criticized by Kant, because they were laid on dogmatic metaphysics. Just as his predecessors, Kant conceived a concept of organism adjusted to its conception which, unlike the previous ones, of a dogmatic character, is marked by a transcendental critical approach, which lies in the relationship of knowledge with the cognitive faculty that produces it the key to understand the 'know' and 'knowledge' of the object of biology. Therefore, the biological knowledge in the Kantian thought finds its deepest and safest ground in the knowing subject, the concept of organism, in other words, of an internal material objective purposiveness (natural purpose), contribution of the subject in the 'composition' of biological knowledge, which is a heuristic rule that guides the reflection before the multiple given data, without which objects such as 'organized beings' could not be judged. The concept of organism is a form of what Kant called 'technique of nature' that makes one think of both nature and objects of nature, as if they were art. This analogical concept provides the rule upon which the faculty of reflecting power of judgement depends on in order to fulfill its task, namely, applying the multiple given data, rising from the particular to the universal. Such ability, by submitting the given multiple data to the concept of 'technique of nature', suggests the determinate nature as a system in accordance with empirical laws and determinate objects as final systems. The organism as an object of nature, is not a simple machine, an aggregate of parts, but a system, whose specific feature is to be the cause and effect of itself (natural purpose), being able to reproduce itself as substance, organ, individual and species. Finally, some considerations are made about the impact of the Kantian concept of organism on the conception of mechanistic biologists from the seventeenth and eighteenth centuries, which compared the organism to the machine. According to Kant, this view makes one think of the organism as an expression of an artificial teleology, whereby which the causality according to ends is thought of as a result from the action of an external author to work, leading the cognitive faculty to a transcendent cause, illegitimate in a critical transcendental approach to the knowledge. However, Kant argues, the organism is an expression of the natural teleology, whereby which the causality according with ends is merely thought of by analogy with art, as if it was an effect of a causality according with ends. Thus, the Kantian concept of organism involves a teleology that pictures itself only as an heuristic principle to cognitive faculty, without which it would not be able to know the peculiarities of the organism
Tissue location of resistance in apple to the rosy apple aphid established by electrical penetration graphs
No full textA study of the constitutive resistance of the apple cultivar Florina, Malus domestica Borkh. (Rosaceae), to the rosy apple aphid, Dysaphis plantaginea (Passerini) (Homoptera Aphididae), was performed for the first time by the electrical penetration graph (DC-EPG) system, using the susceptible apple cultivar Smoothe as control. All experiments were conducted with apterous adult virginoparae. The results showed a constitutive resistance in Florina due to a much longer period before the first probe reflecting surface factors. Some weak indications were found for pre-phloem resistance and initiating phloem access was not affected as inferred from equal time to show phloem salivation. However, the complete absence of phloem ingestion indicates a major resistance factor in the phloem sieve elements, most likely in the sieve element sap. Surface factors could have affected tissue related variables and this should be studied further. Anyhow, the strong constitutive resistance in Florina, either on the surface alone or in the phloem as well, effectively prevented reliable experiments on induced resistance, previously detected by molecular methods
Evaluation of optimization techniques for aggregation
No full textAggregations are almost always done at the top of operator tree after all selections
and joins in a SQL query. But actually they can be done before joins and make later
joins much cheaper when used properly. Although some enumeration algorithms
considering eager aggregation are proposed, no sufficient evaluations are available
to guide the adoption of this technique in practice. And no evaluations are done
for real data sets and real queries with estimated cardinalities. That means it is not
known how eager aggregation performs in the real world.
In this thesis, a new estimation method for group by and join combining traditional
estimation method and index-based join sampling is proposed and evaluated.
Two enumeration algorithms considering eager aggregation are implemented and
compared in the context of estimated cardinality. We find that the new estimation
method works well with little overhead and that under certain conditions, eager
aggregation can dramatically accelerate queries
Evaluation of optimization techniques for aggregation
No full textAggregations are almost always done at the top of operator tree after all selections
and joins in a SQL query. But actually they can be done before joins and make later
joins much cheaper when used properly. Although some enumeration algorithms
considering eager aggregation are proposed, no sufficient evaluations are available
to guide the adoption of this technique in practice. And no evaluations are done
for real data sets and real queries with estimated cardinalities. That means it is not
known how eager aggregation performs in the real world.
In this thesis, a new estimation method for group by and join combining traditional
estimation method and index-based join sampling is proposed and evaluated.
Two enumeration algorithms considering eager aggregation are implemented and
compared in the context of estimated cardinality. We find that the new estimation
method works well with little overhead and that under certain conditions, eager
aggregation can dramatically accelerate queries
Heterogeneous memory technologies in database management systems
No full textDatenbanksysteme sind ein fundamentaler Baustein der modernen IT-Landschaft. Ihr Betrieb trägt häufig deutlich zu den Gesamtkosten von IT-Systemen bei. Folglich besteht hoher ökonomischer Druck die Kosteneffizienz von Datenbanksystemen zu verbessern. Währenddessen ist ein großer Teil der Betriebskosten von Datenbanksystemen au f ihren Hauptspeicherverbrauch zurückzuführen. Verbesserungen der Speichereffizienz führen somit direkt zu höherer Kosteneffizienz. Die zurzeit meistgenutzte grundlegende Speichertechnologie ist DRAM. Allerdings wurden im Laufe der Zeit einige heterogene technologische Alternativen entwickelt. In dieser Arbeit präsentieren wir zwei Ansätze, um die Speichereffizienz von Datenbanksystemen zu verbessern. Zuerst untersuchen wir die Nutzbarkeit von heterogenen Speichertechnologien als ein kosteneffizientere Alternative zu aktuellen Systemen die ausschließlich teuren DRAM-basierten Speicher verwenden. Dazu entwickeln wir ein Kostenmodell zur Platzierung von Daten in Systemen mit hybridem Speicher. Wir stellen fest, dass ein In-Memory Datenbanksystem einen Großteil seiner Basisdaten mit nur geringen Leistungseinbußen in langsameren, günstigeren Speicher platzieren kann. Zweitens führt der konkurrierende Speicherbedarf für Basis- und temporäre Daten in typischen Datenbanksystemen häufig zu einer Überdimensionierung der Speicherkapazität. Somit stellen wir etwas überraschend fest, dass auch ohne den Einsatz heterogener Speichertechnologien noch erhebliche Steigerungen der Speichereffizienz möglich sind. Dazu stellen wir einen kooperativen Ansatz zur Bewältigung von widersprüchlichem Speicherbedarf vor. Wir vergleichen den kooperativen Ansatz mit dem traditionell in bestehenden Systemen verwendeten Ansatz mithilfe eines Prototyps. In unserer Evaluation stellen wir fest, dass kooperative Speicherverwaltung den Gesamtspeicherbedarf eines Datenbanksystems erheblich verringern kann, insbesondere für gemischte Workloads. Letztlich können die beiden in dieser Arbeit skizzierten Wege als solide Grundlage dienen, um die Kosteneffizienz von Datenbanksystemen deutlich zu steigern.Database systems are an important foundation of the modern IT landscape. Their operation often contributes significantly to the total cost of IT systems. Consequently, database systems are under strong economic pressure to improve cost efficiency. A large part of their operating costs is related to main memory consumption. Thereby improving memory efficiency directly leads to higher cost efficiency. The primary underlying technology for main memory today is DRAM, but several heterogeneous technological alternatives have emerged over time. In this work, we present two paths towards higher memory efficiency. First, we explore the viability of using heterogeneous memory technologies as a more cost-efficient alternative to current systems relying exclusively on expensive DRAM-based main memory. To this end, we develop a cost model that can be used to guide data placement in hybrid memory systems. We find that an in-memory database system can place a majority of its base data in slower and cheaper memory with little performance penalty compared to the expensive DRAM-only setup. Second, the conflicting memory needs for base and temporary data in typical database systems often lead to over-provisioned memory capacity. Thus, somewhat surprisingly, we find that significant gains in memory efficiency are still possible even without using heterogeneous memory technologies. We present a cooperative approach for managing these conflicting memory needs, and a prototype system to compare it to the traditional approach used in existing systems. Our evaluation shows that cooperatively managing memory can significantly reduce the overall memory needs of a database system, especially for mixed workloads. Ultimately, the two paths outlined in this thesis can serve as strong foundations to significantly increase the cost efficiency of database systems
- …
