A divide and conquer based distributed run-time mapping methodology for many-core platforms

Anagnostopoulos,; Bartzas,; Kathareios,; Soudris,

doi:10.1109/date.2012.6176442

Cited by 11 publications

(16 citation statements)

References 14 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…Al Faruque et al [1] present a distributed cluster oriented framework for homogeneous platforms which is based on agents for the task-to-cluster mapping. Anagnostopoulos et al [2] present a divide and conquer based distributed run-time mapping framework for both homogeneous and heterogeneous platforms with the introduction of a matching factor. In order to reduce the on-chip node intercommunication Cui et al [4] present a decentralized cluster-based scheme for task mapping, designed for reduction of the communication traffic between agents.…”

Section: Related Workmentioning

confidence: 99%

Distributed run-time resource management for malleable applications on many-core platforms

Anagnostopoulos

Tsoutsouras

Bartzas

et al. 2013

Proceedings of the 50th Annual Design Automation Conference

View full text Add to dashboard Cite

Todays prevalent solutions for modern embedded systems and general computing employ many processing units connected by an on-chip network leaving behind complex superscalar architectures In this paper, we couple the concept of distributed computing with parallel applications and present a workload-aware distributed run-time framework for malleable applications on many-core platforms. The presented framework is responsible for serving in a distributed way and at run-time, the needs of malleable applications, maximizing resource utilization avoiding dominating effects and taking into account the type of processors supporting platform heterogeneity, while having a small overhead in overall inter-core communication. Our framework has been implemented as part of a C simulator and additionally as a runtime service on the Single-Chip Cloud Computer (SCC), an experimental processor created by Intel Labs, and we compared it against a state-of-art run-time resource manager. Experimental results showed that our framework has on average 70% less messages, 64% smaller message size and 20% application speed-up gain.

show abstract

Section: Related Workmentioning

confidence: 99%

Distributed run-time resource management for malleable applications on many-core platforms

Anagnostopoulos

Tsoutsouras

Bartzas

et al. 2013

Proceedings of the 50th Annual Design Automation Conference

View full text Add to dashboard Cite

show abstract

“…with smaller r and M D values. This leads to a higher chance of finding a feasible mapping (Section 5.2.1) along with a better mapping quality (Section 5.2.2) compared to related work [14,29,[33][34][35].…”

Section: Fault-aware Mappingmentioning

confidence: 85%

“…Afterwards, for each M ET task (tc), we collect the set of available accessible nodes ( N -line 9) which are in the first square around AM (r = 1) 3 and with the smallest Manhattan Distance (M D = 1) from parent of tc. Among the collected nodes, we select the one, if any, which keeps the mapping feasible to allocate to tc (lines [10][11][12][13][14][15]. MD increases once all M ET tasks are examined for feasible mappings while current radius, r, increases once all possible MD values are tried.…”

Section: Fault-aware Mappingmentioning

confidence: 99%

“…This is primarily because of the extremely dynamic workload of such systems, where an unpredictable sequence of different applications with various requirements enter and leave the system at run-time. The growing complexity has encouraged several researchers to use distributed resource management methods [11][12][13][14].…”

Section: Introductionmentioning

confidence: 99%

“…Current decentralized approaches [11][12][13][14] completely distribute the resource management job over different nodes. As a result, they lack a pervasive view of the system and might lead to disharmonic and random behaviors [22].…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

SHiFA

Fattah

Palesi

Liljeberg

et al. 2014

Proceedings of the 51st Annual Design Automation Conference

View full text Add to dashboard Cite

A system-level approach to fault-aware resource management of many-core systems is proposed. The proposed approach, called SHiFA, is able to tolerate run-time faults at system level without any hardware overhead. In contrast to the existing system-level methods, network resources are also considered to be potentially faulty. Accordingly, applications are mapped onto healthy nodes of the system at run-time such that their interaction will not require the use of faulty elements. By utilizing the simple routing approach, results show 100% utilizability of PEs and 99.41% of successful mapping when up to 8 links are broken. SHiFA design is based on distributed operating systems, such that it is kept scalable for future many-core systems. A significant improvement in scalability properties is observed compared to the state-of-the-art distributed approaches.

show abstract

On the replacement of objects from round-based applications over heterogeneous environments

Righi

Graebin

Costa

2014

Softw. Pract. Exper.

View full text Add to dashboard Cite

In recent years, there has been growing support for more tightly coupled applications regarding heterogeneous resources. A specific way of obtaining better performance in such applications is to consider the replacement of execution entities by newer resources during the application's lifetime. Therefore, this article describes the rationale for developing jMigBSP, which is a Java programming library that offers object rescheduling for round-based applications. In this context, the proposal addresses Bulk Synchronous Parallel (BSP) applications because BSP represents one of the most often used models for writing tightly coupled parallel programs. jMigBSP's main contribution examines the rescheduling facility in two different ways: (i) using migration directives in the application code directly; and (ii) through automatic load balancing at the middleware level. Specifically, this second idea is feasible because of Java's inheritance feature, which transforms a simple jMigBSP application into a migratable one by changing only a single line of code. In addition to the description of jMigBSP, this article emphasizes the benefits of using migration over heterogeneous environments by executing scientific applications. The results indicate gains of up to 56% with object rescheduling and support the feasibility of using migration as a load balancing technique. The migration of objects to different resources becomes increasingly important for tightly coupled applications because resource availability and application behavior can vary over time on grid environments. The use of a rescheduling technique enables on-the-fly load balancing to reduce the application's time. This dynamic rescheduling could transfer objects from overloaded to underloaded resources or reallocate objects with higher communications patterns in the same cluster. Rescheduling is especially important on long-running applications, which are characterized as CPU intensive, or on large-volume data transfers, because the newly optimized organization can be used at least to amortize the migration penalties [2].Currently, research in the rescheduling field includes the definition of unified metrics for acting in response to application and resource dynamics and user-friendly programming interfaces for providing both application's modeling and replacement facilities [8,9]. Technically, we can benefit from process migration either by modifying the application source code or by using migration-aware communication middleware [10]. The former represents an explicit approach, which usually consists of including migration directives or implementing receiver-initiated load balancing algorithms inside the application's code. The latter is commonly an extension of the programming library for providing a transparent and effortless migration mechanism from the user's point of view. Considering these characteristics, this last approach represents an implicit way of providing migration from the developer's perspective.In this regard, we could ask which is the best approach...

show abstract

A divide and conquer based distributed run-time mapping methodology for many-core platforms

Cited by 11 publications

References 14 publications

Distributed run-time resource management for malleable applications on many-core platforms

Distributed run-time resource management for malleable applications on many-core platforms

SHiFA

On the replacement of objects from round-based applications over heterogeneous environments

Contact Info

Product

Resources

About