nCUBE, Beaverton, Oregon. nCUBE 2 Supercomputers: Technical Overview, 1990.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....only sequential (instead of random) accesses of the data le. From the above discussion, it is not obvious if there is an absolute winner. In order to gain insight on when to use which algorithm, and to determine the most robust technique for general applications, we implemented, on an nCUBE 2 [13] multiprocessor system, four parallel join algorithms representing the four approaches discussed earlier: 1. No Load Balancing (NLB) join algorithm: This scheme is similar to the techniques presented in [9, 15] It represents traditional algorithms which do not perform load balancing at all. 2. ....

....only 12 I O processors. Each processor runs at 7.5 MIPS and has 16 Mbytes of private memory. We pair each processor with a directly connected I O processor and its disk drives to form a PN. With such con guration, the I O and communication bandwidths available to each PN are 2. 2 Mbytes per second [13]. 3.2 Test Data The synthetic database is used as workload because we have a better control over di erent parameters that characterize the data set than the real database. Since we typically perform the join operation between primary key and its foreign key, we focus our attention on the single ....

nCUBE, Beaverton, Oregon. nCUBE 2 Supercomputers: Technical Overview, 1990.

....load balancing until the entire relations have been hashed. From the above discussion, it is not obvious if there is an absolute winner. In order to gain insight on when to use which approach, and to determine the most robust technique for general applications, we implemented, on an nCUBE 2 [17] multiprocessor system, the following four parallel join algorithms which represent the various approaches discussed previously: 1. Conventional Load Balancing (CLB) join algorithm: This scheme is the ABJ algorithm presented in [9] Any one of the bin packing techniques would have been a good ....

....12 processors since our system has only 12 I O processors. Each processor runs at 7.5 MIPS and has 16 MBytes of private 11 memory. We pair each processor with a directly connected I O processor and its disk drives to form a PN. With such a configuration, each PN has a 2. 2 MBytes sec I O bandwidth [17]. 3.2 Test Data A synthetic database is used as the workload because we want to have better control over different parameters that characterize the data set. In practice, join operations are typically performed between a primary key of one relation and a foreign key in another relation. Since ....

nCUBE, Beaverton, Oregon. nCUBE 2 Supercomputers: Technical Overview, 1990.

....disks to separate processors or ports on the interconnection network, as shown in Figure 1. In either case files are declustered over many disks. We call the latter structure Parallel Independent Disks (PID) Examples of multiprocessors using a PID architecture include the Intel [14, 15] nCUBE [22, 5, 27], and Kendall Square Research [20] multiprocessors. Network Memory Processor Memory Processor Memory Processor Disk Disk Disk Figure 1: Parallel Independent Disks (PID) in an MIMD multiprocessor. 3 The workload Parallel file systems and the applications that use them are not sufficiently mature ....

nCUBE Corporation. nCUBE 2 supercomputers: Technical overview. Brochure, 1990.

....to separate processors or ports on the interconnection network, as shown in Figure 2. In either case files are declustered over many disks. We call the latter structure Parallel Independent Disks (PID) Examples of multiprocessors using a PID architecture include the Intel [Int88, Int91] nCUBE [nC90, DdR92, PFDJ89] and Kendall Square Research [KSR92] multiprocessors. 3 The Workload Parallel file systems and the applications that use them are not sufficiently mature for us to know what access patterns might be typical. Here we define our expectations for parallel file access patterns in a ....

nCUBE Corporation. nCUBE 2 supercomputers: Technical overview. nCUBE brochure, 1990.

....processors or ports on the interconnection network. We call the latter structure Parallel Independent Disks (PID) Examples of a PID architecture include Intel s Concurrent File System [15, 8, 17] the Bridge simulated file system [6, 5] for the BBN Butterfly, and the file system for the nCUBE 2 [12, 4, 17]. While caching has not been studied for parallel file systems, Alan Smith has extensively Kotz and Ellis: Caching and Writeback Policies in Parallel File Systems 5 studied caching in uniprocessors with general purpose workloads [20] Uniprocessor and distributed system file access patterns have ....

nCUBE Corporation. nCUBE 2 supercomputers: Technical overview. Brochure, 1990.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC