Scalability Issues in Cluster Computing Operating Systems

 

 

Network of Workstations, popularly known as NOW [6], is quickly becoming a cost-effective and scalable alternative for high-end engineering computing. In large computing environments such as Intel chip design EDA environment, global optimization across several clusters is necessary to satisfy the large scale computing needs of several CPU design teams working on multiple generations of chips. Interconnection of clusters involves different aspects such as resource management, global scheduling, and network bandwidth considerations. In particular it becomes complex when different organizational units support and control different parts of the computing resources. We focused on the topology of the system as the key part of the entire system architecture while addressing the scalability issues. Over the past three years we have been using a multi-cluster batch system and our conclusions are based on this experience. This paper presents different approaches and associated challenges in interconnecting tens of thousands of machines, and shares some of our experiences in scaling and global optimization across multiple clusters. We conclude that multi-level clustering is essential and practical for large scale global distributed computing across several co-operative and geographically dispersed user groups. [via]
http://www.crhc.uiuc.edu/~steve/wcbc99/wcbc-9...

Rating: 0/10

 

Download File

 

Tags: Tag operating system, Tag cluster computing, Tag scalability, Tag issue, Tag scalability issue, Tag distributed computing, Tag job scheduling, Tag load balancing, Tag multi cluster, Tag cluster, Tag batch processing, ...

 

Related Files

 

 
Sponsored Links
Free Download DigiTech Manual, Guide, Instructions, available in PDF ebooks format.

Rate this Document

Popular eBooks

ADS