|
|
Robust and Distribusted OCR Processing System
Raur, Pavel ; Herout, Adam (referee) ; Hradiš, Michal (advisor)
This thesis focuses on creating distributed computing system for document processing using OCR. System is designed to coordinate computation across multiple nodes and distribute tasks between them. Created system was tested for proper functionality. Development is carried out within the PERO project. Developed system will be integrated into the project's demonstration application.
|
|
| |
|
|
Association Attack with Hashcat in a Distributed Environment
Wagner, Lukáš ; Veselý, Vladimír (referee) ; Hranický, Radek (advisor)
The Fitcrack project is a distributed system for cracking cryptographic hashes developed at FIT BUT. The Hashcat tool is used to crack passwords on the computational units. This tool added a new attack mode in 2020 called an association attack. This attack is based on knowledge of a likely password, which is extensively modified during the attack. The goal of this work is to design and implement an extension to the Fitcrack project, which enables the use of the association attack and solves its workload distribution in this distributed environment. Association attack requires modification of distribution methods used by other attacks. Such new methods are proposed and implemented. Implementation is later experimentally verified and conclusion is drawn.
|
|
|
Robust and Distribusted OCR Processing System
Raur, Pavel ; Herout, Adam (referee) ; Hradiš, Michal (advisor)
This thesis focuses on creating distributed computing system for document processing using OCR. System is designed to coordinate computation across multiple nodes and distribute tasks between them. Created system was tested for proper functionality. Development is carried out within the PERO project. Developed system will be integrated into the project's demonstration application.
|
|
|
Balancing Keyword-Based Data and Queries in Distributed Storage Systems
Wirth, Martin ; Parízek, Pavel (advisor) ; Zavoral, Filip (referee)
Research in the area of load balancing in distributed systems has not yet come with an optimal load balancing technique. Existing approaches work primarily with replication and sharding. This thesis overviews existing knowledge in this area with focus on shard- ing, and provides an experiment comparing a state-of-the-art load balancing technique called Weighed-Move with a random baseline and an existing domain-specific balancing implementation. As a significant part of the project, we engineered a generic and scal- able load balancer that may be used in any distributed system and deployed it into an existing ad system called Sklik. The major challenges appeared to be tackling various problems related to data consistency, performance and synchronization, together with solving compatibility issues with the rest of the still-evolving ad system. Our experiment shows that the domain-specific load balancing implementation produces data distribution that enables better performance, but Weighed-Move proved to have a great potential and its results are expected to be enhanced by further work on our implementation. 1
|
|
| |
|
| |
|
|
Optimization of DEECo gossip-based communication
Kováč, Ondrej ; Bureš, Tomáš (advisor) ; Arcaini, Paolo (referee)
The spread of wireless devices inspired the creation of a DEECo component model suitable for designing applications with immanent mobility and dynamic composition where the system architecture emerges at runtime. A great challenge in implementation of such a system is the underlying communication mechanism based on gossip protocol in order to achieve resilience and suitability for MANET networks. In this thesis we propose an optimization of the protocol exploiting infrastructure networks, but still preserving the gossip-like communication without a centralized element. The improvement is based on forming communication groups introduced at the design level. The experiments show a substantial decrease in the number of sent messages and a decrease in time of data delivery. The timing aspect of data delivery is further elaborated for MANET networks by implementing a pulling mechanism with significant improvement of the latency. Part of this thesis is dedicated to a formal specification of the system semantic to provide a precise rationale about its properties and laying the ground for further extensions and research. Powered by TCPDF (www.tcpdf.org)
|
|
|
Parallel and Distributed Processing of Large Textual Data
Straka, Ivan ; Kouřil, Jan (referee) ; Smrž, Pavel (advisor)
This thesis deals with distributed systems, problems related to them, distribution of computing power and load balancing. It describes design and implementation of the distributed system for processing of large textual data, its architecture, loadbalancing, parallel processing of large textual data, communication between nodes, fault detection in communication and maintaining consistency.
|
|
|
Optimization of Distributed Network Flow Collector
Wrona, Jan ; Grégr, Matěj (referee) ; Žádník, Martin (advisor)
This thesis is focused on the optimization of distributed IP flow information collector. Nowadays, the centralized collector is a frequently used solution but is already reaching its performance limits in large scale and high-speed networks. The implementation of the distributed collector is in its early phase and it is necessary to look for solutions that will use it to its full potential. Therefore this thesis proposes a shared nothing architecture without a single point of failure. Using the above proposed architecture, the distributed collector is tolerant to the failure of at least one node. A distributed flow data analysis software, whose performance scales linearly with the number of nodes, is also part of this thesis.
|
guest ::
