Leonardo Jelenkovic

Genetic programming heuristics for multiple machine scheduling

In this paper we present a method for creating scheduling heuristics for parallel proportional machine scheduling environment and arbitrary performance criteria. Genetic programming is used to synthesize the priority function which, coupled with an appropriate meta-algorithm for a given environment, forms the priority scheduling heuristic. We show that the procedures derived in this way can perform similarly or better than existing algorithms. Additionally, this approach may be particularly useful for those combinations of scheduling environment and criteria for which there are no adequate scheduling algorithms.

A novel versatile architecture for Internet of Things

This paper presents an overview of contemporary architectures for Internet of Things and then introduces a simple novel architecture. The main goal of proposed architecture is to remain simple but still applicable to any Internet of Things environment. Specific applications of IoT should be implemented in application layer, using proposed architecture as backbone. Proposed architecture isnt yet fully defined, but ideas on which is based are well defined and should provide straightforward design and implementation.

Simple day-ahead bidding algorithm for a system with microgrids and a distributor

In this paper we present a heuristic distributed algorithm for day-ahead bidding between collection of microgrids that sell their surplus energy to a power grid controller - distributor, who can use that energy with its consumers. Each microgrid is modeled as a node with consumers, energy generator (from renewable sources) and energy storage capabilities. For each time period, distributor negotiate amounts and prices with each microgrid separately, in rounds. Although both sides are modeled, we focus our research on the microgrids bidding algorithm for monetary gain from sold surplus energy. Presented algorithm is simple and produces good results from microgrid perspective. Due to model simplicity, produced results arent always optimal, and convergence had to be additionally helped. In comparison to similar distributed bidding algorithms for systems with storage capacity presented algorithm is simplest and can be used for day-ahead but with presented adaptations also for more complex microgrids and distributor, real-time adjustments and asynchronous negotiations.

Improving thread scheduling by thread grouping in heavily loaded many-core processor systems

Basic scheduling techniques currently employed in operating systems usually dont account for hierarchy of processors cores and caches resulting in suboptimal system efficiency, especially on heavy loaded systems. In this paper we explore possible improvements in scheduling for such heavy loaded many-core systems. Basic idea is to group threads of the same process as close as possible, preferably even on the same processor. We expect that doing so may improve efficiency of processors cache usage, resulting in better overall performance. The idea is tested by adapting the Linuxs simplest scheduler (in code) - a round-robin scheduler. Achieved results are presented in this paper.

Time-constrained programming in Windows NT environment

This paper presents some experiments on Windows NT operating system trying to determine its timing limits. Experiments were concentrated on the context switching abilities in a multithreaded environment. For this purpose a program simulator is written and simulations performed. We detailed some characteristic times for different process priority classes and processors speed. In this way we try to determine the usage of this operating system in a time-constrained (real-time) environment. For comparison, simulations were also performed on Windows 98 and Linux operating systems.

Benu: Operating system increments for embedded systems engineer's education

Most of todays computer systems, including rapidly emerging embedded ones, rely on an operating system. Consequently, the development of embedded systems and related software often requires a deeper understanding of operating systems. This paper presents a new incrementally built operating system and a learning course formed around it. Each increment builds on the previous one and introduces new system elements, new concepts and solutions, and a new set of assignments for improving or extending operations or simply demonstrating its use. Increments and assignments are designed to extend theoretical and practical knowledge in the operating system domain, give experience with non-trivial software systems and their development tools, familiarize the learner with basic computer hardware components and demonstrate device driver construction. The audience targeted by this operating system and course materials includes advanced students with (basic) knowledge of computer architecture, programming and operating systems. In addition, materials may be used individually as part of a lifelong learning process.

Testing Task Schedulers on Linux System

Testing task schedulers on Linux operating system proves to be a challenging task. There are two main problems. The first one is to identify which properties of the scheduler to test. The second problem is how to perform it, e.g., which API to use that is sufficiently precise and in the same time supported on most platforms. This paper discusses the problems in realizing test framework for testing task schedulers and presents one potential solution. Observed behavior of the scheduler is the one used for “normal” task scheduling (SCHED_OTHER), unlike one used for real-time tasks (SCHED_FIFO, SCHED_RR).