Jung-Lok Yu

Resource Co-Allocation : A Complementary Technique that Enhances Performance in Grid Computing Environment

This paper introduces an Availability Check technique (ACT) as a complementary technique to most resource coallocation protocols in the literature. For a given resource co-allocation protocol, ACT tries to reduce the conflicts that happen between co-allocators when they try to allocate multiple resources simultaneously. In ACT, each job checks for the availability state of required resources and gets informed with updates each time one of the resources availability state changes until all the resources become available. Once all required resources become available a job starts applying the given resource co-allocation protocol. Two co-allocation protocols: All-or-Nothing (AONP) and Order-based Deadlock Prevention (ODP2) Protocols are chosen to be the case studies to simulate the proposed technique (ACT). To simulate ACT, each job is allowed to allocate from 1 to 5 different types of resources simultaneously using a uniform distribution. The results show that applying ACT to one of the two protocols outperform the original one. The resource utilization is improved by up to 34% and 41% for AONP and ODP2, respectively. The job response time is improved by up to 13% and 8% for AONP and ODP2, respectively. And the communication overhead is improved by up to 96% and 94% for AONP and ODP2, respectively. Also, applying ACT to AONP represents the most scalable and fully distributed scheme that outperforms original schemes (AONP and ODP2) and competes well with all other schemes presented in this paper.

Design and implementation of an OGSI-compliant Grid broker service

Grid computing promises the ability to share geographically and organizationally distributed resources to increase effective computational power and resource utilization. However, for the Grid computing to be successful, it is very important to provide middleware services that assist Grid users to easily interact with Grid environments. In this paper, we have designed and implemented a new general-purpose OGSI-compliant Grid resource broker service to hide the underlying complexity of the Grid resources from Grid users and to meet not only Grid user requirements but also resource owner policies. It focuses on discovering and scheduling dynamic resources scattered across multiple organizations. Furthermore, it can be integrated with various scheduling services. We also present experimental results and demonstrate the effectiveness of our Grid broker service.

Improving performance of a dynamic load balancing system by using number of effective tasks

Efficient resource usage is a key to achieving better performance in cluster systems. Previously, most research in this area has focused on balancing the load if each node to use the resources of an entire system more effectively. However, we can achieve further improvement in performance when the load balancing system considers the resource requirement according to the task being assigned. This kind of load balancing system, known as an initial job placement system, requires knowledge of the resource usage of a task in order to fit the job to the most suitable node. Since the initial placement requires that the tasks be scheduled before execution, all resource usage must be provided in terms of the prediction. This approach can severely affect the execution time when it uses an inaccurate prediction. We propose a novel load metric termed number of effective tasks in order to resolve the problem arising from inaccurate predictions. Thus, the initial job placement system can work without knowing job resource usage in priori. Simulation results show that the system incurs 11% shorter execution time that the conventional approach using historical behavior-based estimates.

An efficient implementation of Virtual Interface Architecture using adaptive transfer mechanism on Myrinet

User-level communication is investigated by many researchers, in order to resolve the performance degradation of cluster systems due to inefficient communication protocols. It removes the kernel intervention from the critical communication path. Intel, Microsoft and Compaq introduced the Virtual Interface Architecture (VIA), a standard for user-level communication. However, the existing VIA implementation shows low performance in transferring small messages, because it uses a single mechanism to transfer messages without regard to their message size. We implement a high performance VIA, KVIA (Kaist VIA). KVIA, based on descriptor and message size, dynamically selects a proper transfer mechanism. This implementation effectively handles not only large messages but also small messages. Thus, it can be better applied to the systems that frequently use small messages (e.g., lock protocols for software distributed shared memory). The performance of KVIA is reported using round-trip latency and one-way bandwidth. Our results show the round-trip latency of 40 micro-seconds and the maximum one-way bandwidth of 950 Mbits per second, which is about 74% of Myrinet links peak bandwidth.

Impact of exploiting load imbalance on coscheduling in workstation clusters

Implicit coscheduling is known to be an effective technique to improve the performance of parallel workloads in time-sharing clusters. However, implicit coscheduling still does not take into consideration the system behavior like load imbalance that severely affects cluster utilization. In this paper, we propose the use of global information to enhance the existing implicit coscheduling schemes. We also introduce a novel coscheduling approach - named PROC (process reordering-based coscheduling) - based on process reordering exploiting global load imbalance information to coordinate communicating processes. The results obtained from an in-depth simulation study show that our approach significantly outperforms previous ones (by up to 38.4%) by reducing the idle time (by up to 86.9%) and spin time (by up to 36.2%) caused by the load imbalance.

PROC: process reordering-based coscheduling on workstation clusters

Workstation clusters are emerging as a platform for the execution of general-purpose workloads. To use clusters as shared computing servers, scheduling techniques able to effectively handle workloads with diverse characteristics on demands, are required. Implicit coscheduling is known to be an effective technique to improve the performance of parallel workloads in time-sharing clusters. However, implicit coscheduling still does not take into consideration the system behavior like load imbalance that affects cluster utilization. In this paper, we propose the use of global information to enhance the existing implicit coscheduling schemes. We also introduce a novel coscheduling approach based on process reordering exploiting global load imbalance information to coordinate communicating processes. The results obtained from a detailed simulation study show that our approach significantly decreases the average job response time (by up to 21.4%) by reducing the idle time (by up to 55.6%) and spin time (by up to 31.8%) caused by the load imbalance.