Jiri Safarik

The lookahead in a user-transparent conservative parallel simulator

One of the difficulties inherent to conservative parallel discrete event simulation (PDES) is the need for extraction and explicit specification of the lookahead. In this paper we propose a simulator design which allows users to build simulation models with lookahead transparently. This is based on the cooperation between PDES experts and the users, and on the reusability of model components. We introduce the principal design of the simulator and give the results related to its lookahead.

Distributed File System with Online Multi-master Replicas

Mobile devices manage limited resources, above all data storage capacity and data transfer rate. Data amount used by nowadays applications is beyond the capability of mobile devices. A distributed file system presents an appropriate solution, but current distributed file systems are not suitable for mobile devices. We have explored the properties of current distributed file systems in relation to mobile devices. We propose a new distributed file system. The design of the system was verified by its implementation for GNU/Linux operating system at the server side and by the implementation of clients for Microsoft Windows, Windows Mobile, Linux and Android systems.

Security and reliability of distributed file systems

Reliability and security are properties which are required by any computing system. This paper will describe modern trends in distributed file systems (DFS) oriented on reliability and security in these systems. The first part will present the definition of distributed systems and distributed file systems, with a focus on reliability and security in these systems, then describe traditional DFS like AFS, NFS and Coda. The second part of this paper will describe new trends in DFS with a focus on reliability and security.

State of the Art in Distributed File Systems: Increasing Performance

Need of storing huge amounts of data has grown over the past years. The data should be stored for future reuse or for sharing among users. Data files can be stored on a local file system or on a distributed file system. A distributed file system provides many advantages such as reliability, scalability, security, etc. This paper shows new trends in these systems with a focus on increasing performance. These include the organization of data and metadata storage, usage of caching, and design of replication algorithms.

Load Redistribution in Heterogeneous Systems

Today, a number of distributed schedulers exist, but they are centralized or they are too complex and thus require collecting of an important amount of knowledge. In this paper, we present a decentralized method for load distribution, which uses algorithm with modest complexity. In addition, it takes into account the heterogeneity of deployed hardware.

Gradient method with topology discovery for load-balancing in active networks

Active networks bring new possibilities for networking, which can be used to enhance performance of distributed applications. We focus on load balancing and outline new method that at the beginning does not need to know neither virtual topology nor physical one and therefore it can be potentially used for any application on any network topology.

Load-balancing using autonomous co-operating nodes

The existing distributed environments are well suited for time-consuming applications enabling a de-centralized processing. Approaches like active networks and other programmable network designs open the way for new techniques of distributed computing. One of the problems still needed to be addressed is balancing. The paper addresses the specification of the future environment for distributed computing and the need for load-balancing. The structure of the network node is proposed.

Strong password authentication with AKA authentication mechanism

This contribution discusses algorithms for strong authentication of applications in mobile devices. The current LTE and IMS networks provide strong authentication using USIM smart cards based on AKA algorithm. The problem of this authentication is that this authentication is under the sole control of Telco operators. We can expect that more applications will be placed into the IMS environment in the future. These applications will be based either on SIP (video on demand etc.) or HTTP-based protocols (e.g. as government applications or banking applications etc.). They will be provided not only by Telco operators, but also especially by independent third parties — application (content) providers (e.g. government, banks etc.). This contribution proposes new authentication algorithms that combine AKA algorithm with other authentication algorithms. Therefore, the authentication is not under the sole control of Telco operators, still using strong authentication AKA protocol.

Performance improvement of distributed file system using tunneling

The paper deals with the modern concepts of distributed and hierarchical storage systems. Because of the high capacity of modern storage devices, the increasing number of files stored in such systems and the new client access patterns of mobile devices, the architecture of storage systems needs improvement to perform efficiently and respond to the current needs. Our model covers the traditional hierarchical and distributed approaches; as well as algorithms to efficiently organize the inner storage behavior. We evaluate the typical scenarios in the resulting model and show its advantages.

A Contribution to Techniques for Building Dependable Software Systems

Dependability is an essential property of critical systems and it also contributes to the quality of non-critical systems. There already exists an agreement what this term means, but we are still in the process of searching for an answer how to create dependable software systems in a cost-effective way. For different contexts, the how question may have different correct answers. In this paper, we present how an object-capability programming language P, which we have developed, can positively influence the following attributes of dependability: safety, confidentiality, correctness, and robustness. This is demonstrated using P language to build a simple operating system.