An E cient Processor Network Interface for Local Area Multiprocessors
نویسندگان
چکیده
Most computing environments today consist of a number of workstations or personal comput ers PCs connected via a high speed in terconnection network These environments are usually called workstation clusters while acronyms like COWs Clusters of Worksta tions and NOWs Networks of Workstations are popular as well Although worksta tion clusters have the aggregate processing memory and I O capacity to execute high performance applications they usually lack the required hardware support that allows them to present a high performance communication in terface to user applications Network interfaces and communication protocols designed for slow and untrusted networks impose a number of overheads in both message passing and shared memory applications resulting in loss of a large amount of bandwidth at the application al though the underlying network may provide signi cant bandwidth The main sources of these overheads include
منابع مشابه
Scalable Inter-Cluster Communication Systems for Clustered Multiprocessors
As workstation clusters move away from uniprocessors in favor of multiprocessors to support the increasing computational needs of distributed applications, greater demands are placed on the communication interfaces that couple individual workstations. This paper investigates scalable, e cient, and reliable communication systems for multiprocessor clusters that use commodity local area networks ...
متن کاملMultigrain Shared Memory Multigrain Shared Memory
Parallel workstations, each comprising a 10-100 processor shared memory machine, promise cost-e ective general-purpose multiprocessing. This thesis explores the coupling of such smallto medium-scale shared memory multiprocessors through software over a local area network to synthesize larger shared memory systems. Multiprocessors built in this fashion are called Distributed Scalable Shared memo...
متن کاملNetworking in Osprey
Network links have more and more bandwidth while processor frequencies do not increase signi cantly and thus the best way to improve networking performance is to process packets in parallel on multicore machines. This paper describes a networking architecture where incoming network tra c is demultiplexed to user-level network protocol stacks running on di erent cores using a software packet lte...
متن کاملA Low-level Processor Group Membership Protocol for LANS
This paper presents a processor group membership protocol designed to run on top of a local area network. The protocol maintains information about a selected group of stations that explicitly join the protocol by keeping a replica of a global membership table at every member. Additionally, the protocol guarantees that a given station always occupies the same entry in the table. As a result, tab...
متن کاملData Merging for Shared Memory Multiprocessors
Cache coherence, delayed consistency, shared memory multiprocessors We describe an e cient software cache consistency mechanism for shared memory multiprocessors that supports multiple writers and works for cache lines of any size. Our mechanism relies on the fact that, for a correct program, only the global memory needs a consistent view of the shared data between synchronization points. Our d...
متن کامل