Reba 1.2, June 20, 2002 Jim Sibley (jlsibley@us.ibm.com) * Reba is a file system benchmark focused on the dasd driver * and hardware file system using sequential write and read. Since the read * can be deferred to a later time, the read can be timed to * - immediately follow the write, highlighting the effect of memory cache * - after enough data has been written to clear memory, but smaller than * the device cache, highlighting the effect of the DASD cache * - after enough data has been written to clear memory and DASD cache, * highlighting the effect of direct DASD I/O. * * Multiple reba streams can be run simultaneously to study * the bandwidth and effects of parallel I/O. * * The writes are, by default, fsync'd so the write statistics * reflect the actual write to the device. There is an option to turn this off * to show the effects of memory caching. Also, the fsync's are * done after each write to evenly pace the data to the device. * * Reba is based on Bonnie, a file system benchmark which attempts to study * bottlenecks. It is named 'Bonnie' after Bonnie Raitt. Reba has a the two * limitations in studying file system and device behaviour, 1) several of the * the tests are actually CPU tests (get and put character), and 2) the * block read reported is the SECOND read after the data is written, maximizing * the memory caching effect and losing focus on the driver or device. * * Modified 06/20/02, Jim Sibley, (jlsibley@us.ibm.com) * - only output results * * Modified 06/07/02, Jim Sibley, (jlsibley@us.ibm.com) * - fdatasync removed as superfluous * * Modified 05/02/02, Jim Sibley, (jlsibley@us.ibm.com) * - sync changed to fsync/fdatasync to improve parallel performance. * * Modified 01/10/22, Jim Sibley, (jlsibley@us.ibm.com) * - Results to STDOUT, all other to STDERR * - Read or write Information added to final line for sorting * - Nosync as an option (fsync still done on individual files) * * Written, 01/09/02, Jim Sibley, (jlsibley@us.ibm.com) * based on Bonnie.c source, 00/08/30, Kurt Garloff * - Write and/or Read in a single execution * - Writes always followed by an fsync * - file name is an argument - write creates the file, read reads and * existing file * - results report in MB/sec (10k/sec less than a 1% error for 1 MB/sec). * * COPYRIGHT NOTICE: * Copyright (c) for Reba Jim Sibley, 2002 * Copyright (c) for Bonnie Tim Bray, 1990-1996. * * Everybody is hereby granted rights to use, copy, and modify this program, * provided only that this copyright notice and the disclaimer below * are preserved without change. * DISCLAIMER: * Reba has only be tested in a Linux environment. * This program is provided AS IS with no warranty of any kind, and * The author makes no representation with respect to the adequacy of this * program for any particular purpose or with respect to its adequacy to * produce any particular result, and * The author shall not be liable for loss or damage arising out of * the use of this program regardless of how sustained, and * In no event shall the author be liable for special, direct, indirect * or consequential damage, loss, costs or fees or expenses of any * nature or kind. Reba 1.2, June 20, 2002 Jim Sibley (jlsibley@us.ibm.com) To install reba - inspect and change Makfile for target directories for install - make clean - make - make install Sample bash scripts run.test test run of various reba experiments run.256 reba experiment, 1 to 256 MB files, in various combinations reba.clear clears the test volumes reba.comment writes a comment to the experiment log reba.run writes or reads various file sizes to multiple files reba.set set parameters for other reba.xxx bash scripts reba.start.lg start a new experiment log reba8gb simple 8 gb test, writes then reads rebaband simple bandwidth test for parallel I/O