Juan Miguel del Rosario, Rajesh Bordawekar, and Alok Choudhary. "Improved Parallel I/O via a Two-Phase Run-time Access Strategy." In Proceedings of the IPPS '93 Workshop on Input/Output in Parallel 67 Computer Systems, pp. 56--70, Newport Beach, CA, 1993. Also published in Computer Architecture News 21(5), December 1993, pages 31--38.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....buffering while ionodes do it in disk directed I O. There are three parameters for the target of collective buffering: 1) the number and or identity of the target nodes, 2) the amount of target buffer space, and 3) the target data distribution. Collective buffering is best seen in Two Phase I O [dBC93]. In the first phase, data is permuted according to a conforming distribution. This is a permutation of the data across the cnode processes so that it coincides with the underlying file layout (the target data distribution as 12 mentioned above) The actual I O is performed in the second phase. ....

....This is a permutation of the data across the cnode processes so that it coincides with the underlying file layout (the target data distribution as 12 mentioned above) The actual I O is performed in the second phase. Collective buffering has shown orders of magnitude performance improvement [dBC93, MPI96b]. In general, it does not require operating system support since the cnodes are doing the permuting and optimizing of I O requests. Once permuted, cnodes simply make their requests to their designated ionode(s) This makes implementation easier than disk directed I O. Unfortunately, there are some ....

Juan Miguel del Rosario, Rajesh Bordawekar, and Alok Choudhary. "Improved Parallel I/O via a Two-Phase Run-time Access Strategy." In Proceedings of the IPPS '93 Workshop on Input/Output in Parallel 67 Computer Systems, pp. 56--70, Newport Beach, CA, 1993. Also published in Computer Architecture News 21(5), December 1993, pages 31--38.

....perform I O efficiently using such interfaces, compute nodes must request I O in large contiguous blocks, even if a given compute nodes needs just some of the data in a disk block. In such environments the MIOS use a 2 phase approach to I O, first proposed by del Rosario, Bordawekar, and Choudhary [31]. In the 2 phase I O compute nodes read the data using a data distribution that allows compute nodes to issue few I O requests for large blocks of data. The nodes than redistribute the data to conform to the distribution requested by the application. We illustrate how the MIOS module uses 2 phase ....

Juan Miguel del Rosario, Rajesh Bordawekar, and Alok Choudhary, "Improved parallel I/O via a two-phase run-time access strategy," in Proceedings of the IPPS '93 Workshop on Input/Output in Parallel Computer Systems, 1993, pp. 56--70, Also published in Computer Architecture News, 21(5):31-38, December 1993.

....distribution, each processor independently computes the locations of the records it requires from the file, and reads those records. With a datadependent distribution, however, there is no way for processors to request their own set of data. A reasonable solution is similar to two phase I O [11]: each processor reads some convenient subset of data from the file, examines each record to compute the distribution function, and then sends the data to the appropriate processor. In both cases we assume that the distribution function can only decide to which processor each record belongs,and ....

....records. Consider the 64 byte records, which present an interesting picture. Despite nearly doubling the amount of message traffic, traditional caching with data dependent redistribution was faster than the direct access version. Here we were essentially using a pipelined form of two phase I O [11]. Compute processors requested whole blocks from the I O nodes, rather than small records. The larger requests reduced the overhead at the I O nodes, which in this case more than offset the extra work at the compute nodes and in the network. Disk directed I O was always better, however, because it ....

Juan Miguel del Rosario, Rajesh Bordawekar, and Alok Choudhary, "Improved parallel I/O via a twophase run-time access strategy", in IPPS '93 Workshop on Input/Output in Parallel Computer Systems, 1993, pp. 56--70, Also published in Computer Architecture News 21(5), December 1993, pages 31--38.

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