Task Migration Research Articles

Multiprocessor Scheduling of a Multi-Mode Dataflow Graph Considering Mode Transition Delay

The Synchronous Data Flow (SDF) model is widely used for specifying signal processing or streaming applications. Since modern embedded applications become more complex with dynamic behavior changes at runtime, several extensions of the SDF model have been proposed to specify the dynamic behavior changes while preserving static analyzability of the SDF model. They assume that an application has a finite number of behaviors (or modes), and each behavior (mode) is represented by an SDF graph. They are classified as multi-mode dataflow models in this article. While there exist several scheduling techniques for multi-mode dataflow models, no one allows task migration between modes. By observing that the resource requirement can be additionally reduced if task migration is allowed, we propose a multiprocessor scheduling technique of a multi-mode dataflow graph considering task migration between modes. Based on a genetic algorithm, the proposed technique schedules all SDF graphs in all modes simultaneously to minimize the resource requirement. To satisfy the throughput constraint, the proposed technique calculates the actual throughput requirement of each mode and the output buffer size for tolerating throughput jitter. We compare the proposed technique with a method that analyzes SDF graphs in each execution mode separately, a method that does not allow task migration, and a method that does not allow mode-overlapped schedule for synthetic examples and five real applications: H.264 decoder, lane detection, vocoder, MP3 decoder, and printer pipeline.

Migration of long-running Tasks between Reconfigurable Resources using Virtualization

Computing performance and scalability are the essential basics in modern data centres. Field Programmable Gate Arrays (FPGAs) provide a promising opportunity to improve performance, security and energy efficiency. Especially background acceleration of computationally complex and long-running tasks is an important field of application. A flexible use of reconfigurable devices within a cloud context requires an abstraction of the actual hardware through virtualization. In this paper we present an approach inspired by paravirtualized machines for the integration of reconfigurable hardware into cloud services. Using partial reconfiguration our hardware and software framework virtualizes a single physical FPGA to enable multiple independent user designs. Essential components are the management of those virtual user-defined accelerators (vFPGA) and their migration between physical FPGAs to achieve higher system-wide utilization. The migration requires saving and restoring the internal state or context of the vFPGA. We demonstrate the application possibilities and the resource trade-off of our approach by transferring a running design from one physical FPGA to another. Moreover, we present future perspectives for the use of FPGAs in cloud-based environments.

Minimizing scheduling overhead in LRE-TL real-time multiprocessor scheduling algorithm

In this paper, we present a modification of Local Remaining Execution-Time and Local time domain (LRE-TL) real-time multiprocessor scheduling algorithm, aimed at reducing the scheduling overhead in terms of task migrations. LRE-TL achieves optimality by employing the fairness rule at the end of each time slice in a fluid schedule model. LRE-TL makes scheduling decisions using two scheduling events. The Bottom (B) event, which occurs when a task consumes its local utilization thus; it has to be preempted in order to resume the execution of another task, if any, or to idle the processor if none exist. The Critical (C) event occurs when a task consumes its local laxity which means that the task cannot wait any more and has to be scheduled for execution immediately otherwise, it will miss its deadline. Event C always results in a task migration. We have modified the initialization procedure of LRE-TL to make sure that tasks which have higher probability of firing a C event will always be considered for execution first. This will ensure that the number of C events will always be at the minimum; thereby reducing the number of task migrations.

ERAM2 - ENERGY BASED RESOURCE ALLOCATION WITH MINIMUM RECKON AND MAXIMUM RECKON

The emerging field of cloud computing has flexibility and dominant computational architecture that offers ubiquitous services to users. It is different from traditional architecture because it accommodates resources in a unified way. Due to rapid growth in demands for providing the resources and computation in cloud environments, Resource allocation is considered as primary issues in performance, efficiency, and cost. For the provisioning of resource, Virtual Machine (VMs) is employed to reduce the response time and executing the tasks according to the available resources. The users utilize the VMs based on the characteristics of the tasks for effective usage of resources. This helps in load balancing and avoids VMs being in an idle state. Several resource allocation techniques are proposed to maximize the utility of physical resource and minimize the consuming cost of Virtual Machines (VMs). This paper proposes an Energy-Based Resource Allocation with Minimum Reckon and Maximum Reckon (ERAM2); which achieves an efficient scheduling by matching the user tasks on Resource parameters like Accessibility, Availability, Cost, Reliability, Reputation, Response time, Scalability and Throughput in the terms of Maximum Reckon and Minimum Reckon. This paper proposes an Ant Colony - Maximum Reckon and Minimum Reckon (AC-MRMR) method to consolidate all the available resource based on the pheromone value; the score is calculated for each pheromone value. When the score value exceeds Threshold limit then task migration process is carried out for optimized resource allocation of tasks.

Providing fair-share scheduling on multicore computing systems via progress balancing

Fair-share scheduling is essential for achieving performance isolation.Current Linux CFS cannot effectively achieve multicore fairness.A task migration policy, progress balancing, is proposed for multicore fairness.The proposed algorithm extends our previous work via mechanism called throttling.It bounds the virtual runtime difference between any pair of tasks by a constant. Performance isolation in a scalable multicore system is often attempted through periodic load balancing paired with per-core fair-share scheduling. Unfortunately, load balancing cannot guarantee the desired level of multicore fairness since it may produce unbounded differences in the progress of tasks. In reality, the balancing of load across cores is only indirectly related to multicore fairness. To address this limitation and ultimately achieve multicore fairness, we propose a new task migration policy we name progress balancing, and present an algorithm for its realization. Progress balancing periodically distributes tasks among cores to directly balance the progress of tasks by bounding their virtual runtime differences. In doing so, it partitions runnable tasks into task groups and allocates them onto cores such that tasks with larger virtual runtimes run on a core with a larger load and thus proceed more slowly. We formally prove the fairness property of our algorithm. To demonstrate its effectiveness, we implemented our algorithm into Linux kernel 3.10 and performed extensive experiments. In the target system, our algorithm yields the maximum virtual runtime difference of 1.07 s, regardless of the uptime of tasks, whereas the Linux CFS produces unbounded virtual runtime differences.

Dynamic Scheduling in Cloud Computing using Particle Swarm Optimization

Objectives: Dynamic assignment of tasks to different VM in cloud datacenter and Load-Balancing by migration of tasks from an Overloaded VM to Candidate VM. Methods/Statistical Analysis: Particle Swarm Optimization (PSO) algorithm is used finding each iteration the P_BEST (Personal Best Solution) i.e. current solution is compared with the G_BEST (Global Best Solution) i.e. previous best solution and the G_BEST value is updated at each iteration for calculating the minimum execution time. The parameters being calculated are Global_BEST Solution Function, Personal_BEST Solution Function and Average Utilization for each processor. Findings: A Dynamic Approach for Task-Scheduling using the load-balancing technique is implemented in this paper. Two algorithms namely - Particle Swarm Optimization (PSO) and Improved PSO are used and a comparison is made between them based on number of performance parameters like Scheduling Length (Make Span), Total Execution Time and Total Transfer or Migration Time. A Utilization Graph is plotted to show this comparison which compares these algorithms based on their Cloudlet Length (Scheduling Length) and Total Execution Time. Improved PSO algorithm has lesser or minimum execution time as compared to the PSO algorithm because in the Improved PSO Algorithm two parameters are being considered namely- Cloudlet Length and MIPS (Million Instructions Per Second) which leads to maximum utilization of available resources by all VM. Application/Improvements: This approach is used for dynamically assigning tasks to VM and checking maximum utilization of available resources through load balancing and minimizing the overall execution time and migration time. Keywords: Candidate VM, Cloud Computing, Cloudlet, PSO, Task Migration, Transfer Time, Virtual Machine

Quantum entropy source on an InP photonic integrated circuit for random number generation

Random number generators are essential to ensure performance in information technologies, including cryptography, stochastic simulations and massive data processing. The quality of random numbers ultimately determines the security and privacy that can be achieved, while the speed at which they can be generated poses limits to the utilisation of the available resources. In this work we propose and demonstrate a quantum entropy source for random number generation on an indium phosphide photonic integrated circuit made possible by a new design using two-laser interference and heterodyne detection. The resulting device offers high-speed operation with unprecedented security guarantees and reduced form factor. It is also compatible with complementary metal-oxide semiconductor technology, opening the path to its integration in computation and communication electronic cards, which is particularly relevant for the intensive migration of information processing and storage tasks from local premises to cloud data centres.

Optimum Software Aging Prediction and Rejuvenation Model for Virtualized Environment

<p><em>Advancement in electronics and hardware has resulted in multiple softwares running on the same hardware. The result is multiuser, multitasking and virtualized environments. However, reliability of such high performance computing systems depends both on hardware and software. For hardware, aging can be dealt with replacement. But, software aging needs to be dealt with software only. For aging detection, a new approach using machine learning framework has been proposed in this paper. For rejuvenation, Adaptive Genetic Algorithm (A-GA) has been developed to perform live migration to avoid downtime and SLA violation. The proposed A-GA based rejuvenation controller (A-GARC) has outperformed other heuristic techniques such as Ant Colony Optimization (ACO) and best fit decreasing (BFD) for migration. Results reveal that the proposed aging forecasting method and A-GA based rejuvenation outperforms other approaches to ensure optimal system availability, minimum task migration, performance degradation and SLA violation.</em></p>

Survey and future directions of fault-tolerant distributed computing on board spacecraft

Current and future space missions demand highly reliable on-board computing systems, which are capable of carrying out high-performance data processing. At present, no single computing scheme satisfies both, the highly reliable operation requirement and the high-performance computing requirement. The aim of this paper is to review existing systems and offer a new approach to addressing the problem. In the first part of the paper, a detailed survey of fault-tolerant distributed computing systems for space applications is presented. Fault types and assessment criteria for fault-tolerant systems are introduced. Redundancy schemes for distributed systems are analyzed. A review of the state-of-the-art on fault-tolerant distributed systems is presented and limitations of current approaches are discussed. In the second part of the paper, a new fault-tolerant distributed computing platform with wireless links among the computing nodes is proposed. Novel algorithms, enabling important aspects of the architecture, such as time slot priority adaptive fault-tolerant channel access and fault-tolerant distributed computing using task migration are introduced.

Hierarchical Dynamic Thermal Management Method for High-Performance Many-Core Microprocessors

It is challenging to manage the thermal behavior of many-core microprocessors while still keeping them running at high performance since the control complexity increases as the core number increases. In this article, a novel hierarchical dynamic thermal management method is proposed to overcome this challenge. The new method employs model predictive control (MPC) with task migration and a DVFS scheme to ensure smooth control behavior and negligible computing performance sacrifice. In order to be scalable to many-core systems, the hierarchical control scheme is designed with two levels. At the lower level, the cores are spatially clustered into blocks, and local task migration is used to match current power distribution with the optimal distribution calculated by MPC. At the upper level, global task migration is used with the unmatched powers from the lower level. A modified iterative minimum cut algorithm is used to assist the task migration decision making if the power number is large at the upper level. Finally, DVFS is applied to regulate the remaining unmatched powers. Experiments show that the new method outperforms existing methods and is very scalable to manage many-core microprocessors with small performance degradation.

Theorem proving based Formal Verification of Distributed Dynamic Thermal Management schemes

Distributed Dynamic Thermal Management (DDTM) schemes are widely being used nowadays to cater for the elevated chip temperatures for many-core systems. Traditionally, DDTM schemes are analyzed using simulation or emulation but the non-exhaustive and incomplete nature of these analysis techniques may compromise on the reliability of the chip. Recently, model checking has been proposed for formally verifying simple DDTM schemes but, despite several abstractions, the analysis is limited to less than 100 cores due to the state-space explosion problem. As a more scalable approach for next-generation many-core systems, we propose a methodology based on theorem proving to perform formal verification of DDTM schemes. The proposed approach allows specification and verification of both functional and timing properties for any number of cores and for all times. For this purpose, the paper provides a higher-order-logic formalization of a generic DDTM scheme. The proposed generic model can be specialized to formally specify most of the existing DDTM schemes and thus formally verify their thermal properties, like temperature bounds and balancing and time to reach thermal stability, as higher-order-logic theorems. As an illustrative example, the paper presents a formal model and analysis of a Distributed Task Migration based DDTM scheme for many-core systems.

New approach based on group technology for the consolidation problem in cloud computing-mathematical model and genetic algorithm

Resource management is a hotspot issue in distributed systems like cloud computing (CC). It means how to prepare the computational resources, i.e., servers and virtual machines (VMS), to execute the tasks. This paper offers a new approach based on Group Technology (GT)—known as a powerful philosophy for the resource management in cellular manufacturing systems—to deal with the resource management problem in CC. We develop a mathematical model to optimally consolidate the VMs, servers and tasks simultaneously to control several important factors such as task migrations and server load variation, as well as the number of VMs. To test the validity of our proposed model, several small problems are generated randomly and solved by LINGO 9 software. Furthermore, to cope with larger problems, which cannot be solved optimally, a genetic algorithm is proposed. We, finally, compare our methods with the most well-known algorithms in this context, round robin (RR) and first-come, first-served (FCFS) algorithms.

Performance/energy aware task migration algorithm for many‐core chips

This study proposes an efficient task migration algorithm for mesh-based multi- and many-core chips. The proposed algorithm collects tasks running on a rectangular-based set of cores, that is, source sub-mesh and moves the tasks to another rectangular-based set to remove chip temperature hotspots and to provide balanced load on the chip. The proposed migration algorithm uses the concept of gathering/scattering to minimise the traffic induced by the migration. In this regard, the proposed algorithm uses a selected node in each row of the source sub-mesh to gather tasks of all cores in the same row. Selection of the gathering node is done based on its location in the row and traffic rate of other cores in the row. When gathering nodes are migrated, in the destination sub-mesh, then, they scatter their tasks according to the same pattern among the cores in their rows. Simulations of the proposed migration algorithm are done by Access Noxim simulator in a various range of network conditions with application graphs of D263DECMP3DEC, DMPEG4, and DVOPD. Results obtained from simulations show that the proposed algorithm offers 36% better performance, 28% lower energy consumption, and 7% lower temperature in comparison with the previously proposed migration algorithms.

FAMe-TM: Formal analysis methodology for task migration algorithms in Many-Core systems

Distributed Dynamic Thermal Management (dDTM) through task migrations across cores provides a very promising solution to cater for the heating issues in Many-Core architectures. However, the growing number of cores, the distributed nature of dDTM and the inherent sampling-based nature of traditional analysis techniques, like simulation and emulation, makes a complete and rigorous analysis of these task migration algorithms almost impossible. These limitations compromise the analysis integrity and in worst cases may lead to the deployment of an inefficient and inaccurate dDTM scheme on chip, which in turn can cause permanent defects in the chip due to excessive heating. Leveraging upon the exhaustive nature of model checking based verification, we propose to use a model checker to formally verify task migration algorithms. This work proposes an analysis methodology, i.e., Formal Analysis Methodology for Task Migrations (FAMe-TM), and identifies a generic set of properties for the formal verification of task-migration-based dDTM schemes. In particular, we propose an analysis flow using the scalable bounded model checker, nuXmv, to formally verify the suggested task migration properties, like tasks migrations, stalls, completion, creation of hot spots, time spent in migration and time to achieve stability. For illustration purposes, we apply FAMe-TM to two recently proposed task-migration-based dDTM schemes, i.e., Thermal Coupling Aware (TCA-TM) dDTM and Hot Spot Reduction (HR-TM) dDTM. • A formal analysis methodology, FAMe-TM, for task migration (TM) algorithms is given. • Distributed Dynamic Thermal Management (dDTM) schemes are analyzed using FAMe-TM. • A set of properties are proposed and formalized for comparing TM-based dDTM schemes. • An improved version of a dDTM scheme is proposed based on the findings using FAMe-TM.

Parallel extremal optimization in processor load balancing for distributed applications

The paper concerns parallel methods for extremal optimization (EO) applied in processor load balancing in execution of distributed programs. In these methods EO algorithms detect an optimized strategy of tasks migration leading to reduction of program execution time. We use an improved EO algorithm with guided state changes (EO-GS) that provides parallel search for next solution state during solution improvement based on some knowledge of the problem. The search is based on two-step stochastic selection using two fitness functions which account for computation and communication assessment of migration targets. Based on the improved EO-GS approach we propose and evaluate several versions of the parallelization methods of EO algorithms in the context of processor load balancing. Some of them use the crossover operation known in genetic algorithms. The quality of the proposed algorithms is evaluated by experiments with simulated load balancing in execution of distributed programs represented as macro data flow graphs. Load balancing based on so parallelized improved EO provides better convergence of the algorithm, smaller number of task migrations to be done and reduced execution time of applications.

A survey on decision making for task migration in mobile cloud environments

Mobile cloud computing (MCC) is a new converged computing paradigm. The central idea of MCC is to use powerful backend computing nodes to enhance capabilities of small devices. An effective approach to realize this is to use task migrations. There are already a number of research efforts to MCC helping to make decisions for task migrations. In this paper, we make a short survey on decision making for task migrations in MCC. From the survey we found that there are still some challenges such as comprehensive context-awareness, migration standards, and large scale experiments and good algorithms are needed to support task migration. The objective of this paper is to highlight these issues and challenges to attract more efforts to these questions.

Distributed Sensor Network-on-Chip for Performance Optimization of Soft-Error-Tolerant Multiprocessor System-on-Chip

As transistor density continues to increase with the advent of nanotechnology, reliability issues raised by more frequently appeared soft errors are becoming even more critical to the next-generation multiprocessor systems. In this paper, we present a systematic approach to address the soft-error problem in multiprocessor system-on-chip with the consideration of system performance optimization. To guarantee the system correctness, a hardware–software collaborated approach is proposed to protect the processors from soft errors. Tiny hardware sensors are embedded in the processor cores to detect the soft errors, and the software-based rollback scheduling mechanisms are applied for error recovery. The protection costs on hardware duplication and software redundancy are effectively reduced. To optimize the system performance, a distributed control system is built on top of the on-chip communication network and collaboratively manages the entire chip for application execution. With the cluster-based task migration techniques, an efficient runtime task remapping and rescheduling algorithm is proposed to further mitigate the overheads induced by soft-error protection and to minimize the total performance degradation. The distributed control strategy makes the system more adaptable and flexible to the development of the next-generation hardware and software with larger scales. Extensive performance evaluations using SystemC-based cycle-accurate simulations on a set of real-world applications show that our approach has on average 49% performance improvement and 79.6% energy consumption reduction compared with the related state-of-the-art techniques, and hardware synthesis results show that our approach only introduces 2.9% chip area overheads.

MPSoC application resilience by hardware-assisted communication virtualization

Future MPSoCs will become increasingly vulnerable to environmental and manufacturing variabilities, which can lead to transient and permanent faults during operation. These dependability issues can be tackled with a system-level approach, which migrates tasks away from unreliable processor elements. In this paper we evaluate the usage of protection switching techniques to enable an efficient communication reconfiguration during task migration, which minimizes the task downtime and increases the applicability of this approach. We compare the performance of the dualcast and forward protocols with the common approach of a communication suspension during migration. Furthermore, we investigate the offload of the protocol handling from the productive processor cores and the differences with stateless and stateful task migration. The results show that dualcast and forward outperform the common approach in terms of communication latencies and enable a nearly seamless migration of stateless tasks if the protocol is offloaded.

Mobility Aware Task Allocation for Mobile Cloud Computing

The Mobile Cloud Computing is a promising technology that has provided a way to overcome the limitations of the mobile devices. The advancement of mobile devices technology has made the applications of these devices more complex and resource famished. Mobile cloud computing has created opportunities to execute these applications on the mobile devices by migrating the compute intensive task to the cloud. This migration of task to the cloud is not an easy task. The connectivity of the devices and the cloud is affected by the network inconsistency of wireless network. The servers on the cloud are heterogeneous in nature. Furthermore, the users are most of the time in mobile state which results in frequent change in association to access points. All of these make the selection of an optimal server to offload the task in cloud into a challenging work. In this paper, a comparative survey is provided for allocating task on the cloud along with their limitation. A mobility aware task allocation system for mobile cloud computing is also proposed. An optimization problem is formulated considering the workload and service rate of servers, network inconsistency, time to execute the task, mobility of the users etc. The proposed system aims to allocate task to the server where minimum response time is achieved in order to enhance users’ quality of experience.

Performance evaluation of task migration in contiguous allocation for mesh interconnection topology

In this paper, several efficient migration algorithms have been proposed to improve existing non-preemptive sub-mesh allocation strategies in multiprocessors which are two-row boundary migration algorithm (TRBMA) and different types of combined migration mechanisms. These methods are presented to solve the external fragmentation problem in contiguous allocation. In this process, the mentioned methods contain the algorithms, called the minimum threshold and prime number (MT&MPN), which determine the best sub-mesh size for the incoming job. These algorithms are used to increase continuity in multiprocessor systems. Using extensive simulations, we evaluated the proposed methods and compared them with previous strategies. Simulation outcomes clearly showed the TRBMA and the MT&MPN procedure produced the best average job execution time and average system utilization. Also, hybrid migration methods produced the best average job waiting time and average number of waiting jobs in comparison with the previous algorithms. Meanwhile, most of the traditional methods had better average job response time.