Máté J. Csorba

Ant system for service deployment in private and public clouds

Large-scale computing platforms that serve thousands or even millions of users through the Internet are on a path to become a pervasive technology available to companies of all sizes. However, existing technologies to enable this kind of scaling are based on a hierarchically managed approach that does not scale equally well. Moreover, existing systems are also not equipped to handle the dynamism that may emerge as a result of severe failures or load surges. In this paper, we conjecture that using self-organizing techniques for system (re)configuration can improve both the scalability properties of such systems as well as their ability to tolerate churn. Specifically, the paper focuses on deployment of virtual machine images onto physical machines that reside in different parts of the network. The objective is to construct balanced and dependable deployment configurations that are resilient. To accomplish this, a method based on a variant of Ant Colony Optimization is used to find efficient deployment mappings for a large number of virtual machine image replicas that are deployed concurrently. The method is completely decentralized; ants communicate indirectly through pheromone tables located in the nodes. An example scenario is presented and simulation results are obtained for the method.

Cost-efficient deployment of collaborating components

We study the problem of efficient deployment of software components in a service engineering context. Run-time manipulation, adaptation and composition of entities forming a distributed service is a multi-faceted problem challenged by a number of requirements. The methodology applied and presented can be viewed as an intersection between systems development and novel network management solutions. Application of heuristics, in particular artificial intelligence in the service development cycle allows for optimization and should eventually grant the same benefits as those existing in distributed management architectures such as increased dependability, better resource utilization, etc. The aim is finding the optimal deployment mapping of components to physically available resources, while satisfying all the non-functional requirements of the system design. Accordingly, a new component deployment approach is introduced utilizing distributed stochastic optimization.

Adaptable model-based component deployment guided by artificial ants

We investigate a means for efficient deployment of distributed services comprising of software components. Our work can be viewed as an intersection between model-based service development and novel network management architectures. In a service engineering context, models of services embellished with non-functional requirements are used as input to our swarm intelligence based deployment logic. Mappings between resources provided by the execution environment and components are the results of our heuristic optimization procedure that takes into account requirements of the services. Deployment mappings will be used as feedback towards the designer and the provider of the service. Moreover, our heuristic algorithm possesses significant potential in adaptation of services to changes in the environment.

Foraging for Better Deployment of Replicated Service Components

Our work focuses on distributed software services and their requirements in terms of system performance and dependability. We target the problem of finding optimal deployment mappings involving multiple services, i.e. mapping service components in the software architecture to the underlying platforms for best possible execution. We capture important non-functional requirements of distributed services, regarding performance and dependability. These models are then used to construct appropriate cost functions that will guide our heuristic optimization method to provide better deployment mappings for service components. This paper mainly focuses on dependability. In particular, a logic enabling replication management and deployment for increased dependability is presented. To demonstrate the feasibility of our approach, we model a scenario with 15 services each with different redundancy levels deployed over a 10-node network. We show by simulation how the deployment logic proposed is capable to satisfy replica deployment requirements.

warm intelligence heuristics for component deployment

We address the problem of efficient deployment of software services into a networked environment. Services are considered that are provided by collaborating components. The problem of obtaining efficient mappings for components to host in a network is challenged by multiple dimensions of quality of service requirements. In this paper we consider execution costs for components and communication costs for the collaborations between them. Our proposed solution to the deployment problem is a nature inspired distributed heuristic algorithm that we apply from the service providers perspective. We present simulation results for different example scenarios and present an integer linear program to validate the results obtained by simulation of our algorithm.

A Bio-inspired Method for Distributed Deployment of Services

We look at the well-known problem of allocating software components to compute resources (nodes) in a network, given resource constraints on the infrastructure and the quality of service requirements of the components to be allocated to nodes. This problem has many twists and angles, and has been studied extensively in the literature. Solving it is particularly problematic when there is extensive dynamism and scale involved. Typically, heuristics are needed.In this paper, we present a new breed of heuristics for solving this problem. The distinguishing feature of our approach is a decentralized optimization framework aimed at finding near optimal mappings within reasonable time and for large scale. Three different incarnations of the problem are explored through simulations. For one problem instance, we also provide exact solutions, and show that our technique is able to find near optimal solutions with low variance. In the largest example, a public-private cloud computing scenario is used, where different clouds are associated with financial costs, and we show that our approach is capable of balancing the load as expected for such a scenario.

Component deployment using parallel ant-nests

This paper targets the problem of efficient software component deployment in distributed networked services. We approach the problem from a model-based service development aspect and propose a solution based on novel methods applied in network management. The swarm intelligence-based deployment logic we have developed uses service models that are defined in UML 2.0. These service models are embellished to contain the non-functional requirements against the implementation of the service that are necessary to solve the component deployment problem and to obtain a deployment mapping. The result of the heuristic optimisation procedure is mappings between the components and the resources provided by the execution environment the service is deployed into. In this work, the deployment is viewed from the perspective of the service provider and our algorithm possesses significant potential in providing adaptation support for services in various changing environments.

Laying Pheromone Trails for Balanced and Dependable Component Mappings

This paper presents an optimization framework for finding efficient deployment mappings of replicated service components (to nodes), while accounting for multiple services simultaneously and adhering to non-functional requirements. Currently, we consider load-balancing and dependability requirements. Our approach is based on a variant of Ant Colony Optimization and is completely decentralized, where ants communicate indirectly through pheromone tables in nodes. In this paper, we target scalability; however, existing encoding schemes for the pheromone tables did not scale. Hence, we propose and evaluate three different pheromone encodings. Using the most scalable encoding, we evaluate our approach in a significantly larger system than our previous work. We also evaluate the approach in terms of robustness to network partition failures.