Krishchenko

The paper describes a method of tracking individual IP packets throughout the network from their creation to delivery. The method keeps tracking various events including fragmentation, tunneling and NAT. An experiment with a software implementation of the method is described. Possible applications of the method include troubleshooting and studying of computer networks.

During testing finite discrete models to avoid constraints of available memory, parallel generation and distributed storage of state space can be used. The efficiency of such an approach depends on the choice of the state distribution function between computational nodes. The article considers the choice of this function, which allowed reducing the validation time at the expense of the number of removed callings between the nodes. The experiment results obtained with the help of the prototype of the tool for model validation are included in the article.  

The authors describe problems with the current LRU-based page reclaiming algorithm used by Linux kernel that arise in the case of a regular memory access pattern. Basing on the results of some preliminary experiments, a pattern detection method is proposed. The method uses page fault history as an input. The developed reclaiming algorithm and its implementation for Linux kernel are described. Experiment results with the method software implementation of this algorithm are presented.

The paper discusses a method of retrieving detailed information on IP packet processing in Linux operating system. Modifications to Linux networking subsystem source code are described, along with the architecture of a monitoring tool based on the method being discussed. An example of the monitoring tool output is provided.  Possible applications of the method include troubleshooting and studying computer networks.

The authors consider distributed fault-tolerant DBMS based on the Erlang runtime environment. As the result of testing, it was determined that the bundled software allowed achieving almost a linear growth of speed of query execution depending on the number of replication clusters. Recommendations on system deployment were produced.