Eryk Laskowski

Extremal Optimization applied to load balancing in execution of distributed programs

The paper describes methods for using Extremal Optimization (EO) for processor load balancing during execution of distributed applications. A load balancing algorithm for clusters of multicore processors is presented and discussed. In this algorithm the EO approach is used to periodically detect the best tasks as candidates for migration and for a guided selection of the best computing nodes to receive the migrating tasks. To decrease the complexity of selection for migration, the embedded EO algorithm assumes a two-step stochastic selection during the solution improvement based on two separate fitness functions. The functions are based on specific models which estimate relations between the programs and the executive hardware. The proposed load balancing algorithm is assessed by experiments with simulated load balancing of distributed program graphs. The algorithm is compared against a greedy fully deterministic approach, a genetic algorithm and an EO-based algorithm with random placement of migrated tasks.

Program Design Environment for Multicore Processor Systems with Program Execution Controlled by Global States Monitoring

A new distributed program graphical design environment is described in the paper. It is oriented towards designing program execution control based on a built-in system infrastructure which enables easy global application states monitoring in systems based on multicore processors. Two aspects of global application control design are covered. First is the global control flow in programs at the level of processes and threads. The second is the asynchronous control of internal process and thread behavior. The proposed control infrastructure is based on structural program elements called synchronizers organized at the process and thread levels to collect state information, evaluate control predicates on global states and send signals to application program threads and processes to stimulate global control actions. The paper presents principles of the application program graphical design and programming methods to implement global control at the level of threads.

Byte-code scheduling of Java programs with branches for desktop grid

A method for an introductory optimization of multithreaded Java programs for execution on clusters of Java Virtual Machines (JVMs) inside desktop grids is presented. It is composed of two stages. In the first stage, a clustering algorithm is applied to extended macro data flow graphs generated on the basis of the byte-code compiled for multithreaded Java programs. These graphs account for data and control dependencies in programs including conditional branch instructions annotated by branch statistics driven from execution traces for representative sets of data. In the second stage, a list scheduling is performed based on the Earliest Task First (ETF) heuristics in which node mapping on JVMs accounts for mutually exclusive paths outgoing from conditional branch nodes. The presented object placement optimization algorithm is a part of the DG-ADAJ environment.

Load balancing in distributed applications based on extremal optimization

The paper shows how to use Extremal Optimization in load balancing of distributed applications executed in clusters of multicore processors interconnected by a message passing network. Composed of iterative optimization phases which improve program task placement on processors, the proposed load balancing method discovers dynamically the candidates for migration with the use of an Extremal Optimization algorithm and a special quality model which takes into account the computation and communication parameters of the constituent parallel tasks. Assessed by experiments with simulated load balancing of distributed program graphs, a comparison of the proposed Extremal Optimization approach against a deterministic approach based on a similar load balancing theoretical model is provided.

New program structuring heuristics for multi-processor systems with redundant communication resources

The paper presents a new graph structuring algorithm for look-ahead reconfigurable multi-processor systems. An application program in such a kind of system is partitioned into sections, which are executed using redundant communication resources. Thus program scheduling in a look-ahead environment incorporates the graph partitioning problem. The new proposed algorithm is based on list scheduling and utilizes a new section clustering heuristics for graph partitioning. The main difference between the proposed heuristics and existing ones lies in the new partitioning methodology, based on the critical path controlled, iterative algorithm. The new heuristics for vertex selection are also applied, which use the extensive range of the system and program parameters, including the reconfiguration critical point function. The experimental results are presented and compared with results of existing graph partitioning algorithms for such a kind of environment.

Optimizing Distributed Data Mining Applications Based on Object Clustering Methods

The exponential computational cost involved in traditional data mining methods enforces search for less complex new algorithms. Especially, data mining on grid is a challenge due to the lack of shared memory in grid computing, which puts special attention to communication optimization. The aim of the DisDaMin project (distributed data mining), described in the paper, is solving data mining problems by using new distributed algorithms intented for execution in grid environments. The DisDaMin implements intelligent fragmentation of data by clustering methods and asynchronous collaborative processing adjusted to grid environments. The DG-ADAJ environment provides adaptive control of distributed applications written in Java for desktop grid. It constitutes a component-based middleware, which allows for optimized distribution of applications on clusters of Java virtual machines, monitoring of application execution and dynamic online balancing of processing and communication. The DG-ADAJ system provides a middleware platform for desktop grid that could be used as a deployment base for DisDaMin algorithms. In this paper, we propose static object placement optimization algorithms for fragmentation of data in the DisDaMin project. The algorithms use DG-ADAJs object clustering methods to provide optimized local processing on each node with minimized inter-node communication

Extremal optimization approach applied to initial mapping of distributed java programs

An extremal optimization algorithm for initial Java program placement on clusters of Java Virtual Machines (JVMs) is presented. JVMs are implemented on multicore processors working under the ProActive Java execution framework. Java programs are represented as Directed Acyclic Graphs in which tasks correspond to methods of distributed active Java objects that communicate using a RMI mechanism. The presented probabilistic extremal optimization approach is based on the local fitness function composed of two sub-functions in which elimination of delays of task execution after reception of required data and the imbalance of tasks execution in processors are used as heuristics for improvements of extremal optimization solutions. The evolution of an extremal optimization solution is governed by task clustering supported by identification of the dominant path in the graph. The applied task mapping is based on dynamic measurements of current loads of JVMs and inter-JVM communication link bandwidth. The JVM loads are approximated by observation of the average idle time that threads report to the OS. The current link bandwidth is determined by observation of the performed average number of RMI calls per second.

Extremal optimization applied to task scheduling of distributed Java programs

The paper presents new Java programs scheduling algorithms for execution on clusters of Java Virtual Machines (JVMs), which involve extremal optimization (EO) combined with task clustering. Two new scheduling algorithms are presented and compared. The first employs task clustering to reduce an initial program graph and then applies extremal optimization to schedule the reduced program graph to system resources. The second algorithm applies task clustering only to find an initial solution which is next improved by the EO algorithm working on the initial program graph. Both algorithms are also compared to an EO algorithm which does not use the clustering approach.

A Framework for Desktop GRID Applications: CCADAJ

We propose a new component framework over a middleware platform for grid computing: DG-ADAJ (desktop grid-adaptive distributed application in Java). Our platform allows having a single system image (SSI) of the grid platform. It gives a special mechanism at middleware level which assures dynamic and automatic adaptation to variations of computation methods and execution platform. DG-ADAJ gives also specific mechanisms based on control components which helps a users to build component-based parallel/distributed applications. We use the CCA (common component architecture) as a component architecture model for our framework. CCADAJ (CCA-ADAJ) is a layer above the DG-ADAJ environment, which helps user to build his parallel/distributed applications by assembling components in a transparent way and which exploits the parallelism of the environment

Efficient Parallel Embedded Computing through Look-Ahead Configured Dynamic Inter-Processor Connections

A new kind of the look-ahead reconfigurable multiprocessor embedded systems together with the necessary program structuring algorithms are presented and discussed. The new system architecture for execution of dedicated parallel programs is based on the look-ahead dynamic reconfiguration of inter-processor connections in redundant network resources such as multiple connection crossbar switches and partitioned processor communication link sets. The discussed approach enables reducing inter-processor connection reconfiguration time overheads, which is particularly important in modern parallel systems based on user-level communication. Automatic program structuring methods for the assumed system architecture are proposed. Experimental results with structuring of parallel numerical programs (Strassen matrix multiplication) are reported. The experiments compare the strategies of the look ahead connection reconfiguration based on the use of multiple crossbar switches with the strategy of providing processors with larger sets of communication links partitioned into a subset used for current program execution and a subset used for anticipated connection setting