PhD Thesis

  1.  

    Melashchenko Andrii. Tools and methods for interoperability of national systems of digital signatures. - Manuscript

    Thesis for Ph.D. degree in physics and mathematics by specialty 01.05.03 - Mathematical and software of computers and systems. - VM Glushkov Institute of Cybernetics of the National Academy of Sciences of Ukraine, Kiev, 2010.

    Identified the main problems of interoperability of NSES through a comparison of it with a reference QPKI model. Proposed standard representation of the signing policy as a basic component of the business model of QPKI according to Directive 1999/93/EC. Proposed security profiles of signature suite GOST 34.311 + DSTU 4145 for interoperability of its implementation in cryptomodules. Implemented security profile for signature suite GOST 34.311 + DSTU 4145 as a crytomodule «UPGCryptoProviderBasic» which was implemented according to the Microsoft CryptoAPI specifications. Was proposed formal procedure of accreditation of CA, based on national standards of Ukraine, harmonized with the European. Developed and implemented the specification testbed for assessing the interoperability of implementations of the standard model QPKI on the basis of a formal accreditation procedure CA. Was proposed directions of development of public key infrastructure, based on the analysis of the homomorphic transform notation in the ASN.1-XSD-schema XML-document.

  2. Modern high performance cluster computers being extremely scalable and rather cheap are successfully used for processing of large volumes of data in a many research and industrial areas. A lot of the problems solved by such computers are related to uniform data processing. This includes data mining, simulation, exploration and prospection problems. High efficiency, however, is not always achieved especially on large clusters with hundreds of CPUs. The purpose of the thesis is to analyze efficiency of uniform data processing on cluster computers and develop a technique for predicting performance of such processing on a cluster based on it’s technical specification.

    The result of the research is aset of estimates for a different patterns of parallel programming(Mattson et al.) and data storage and delivery strategies based on queueing theory. Such estimates can help in the parallel program development planning and the data distribution optimization. Besides, the formulas can support decisions on planning the hardware upgrades or selecting more efficient cluster architecture under the budget limitations. A crosspoint of the weak and full load diagrams can be used as a first approximation of the best process number for the specific program run. Using the estimates a technique was developedto provide an evaluation of parallel software performance on an arbitrary cluster given it’s processor, storage and network specification by conducting several experiments on a single or multiple processor basic computer.

    The technique was successfully used for evaluating performance of automatic markup of text corpora, genetic algorithm for data clustering, seismic wave modeling and seismic data migration. As the result the technique was embedded in cluster management system ACMS for series of Inparcom cluster computers, that was developed by V.M. Glushkov Institute of Cybernetics together with Government scientific and industrial organization Electronmash.