1 .\" Copyright (c) 1983 The Regents of the University of California.
2 .\" All rights reserved.
4 .\" Redistribution and use in source and binary forms, with or without
5 .\" modification, are permitted provided that the following conditions
7 .\" 1. Redistributions of source code must retain the above copyright
8 .\" notice, this list of conditions and the following disclaimer.
9 .\" 2. Redistributions in binary form must reproduce the above copyright
10 .\" notice, this list of conditions and the following disclaimer in the
11 .\" documentation and/or other materials provided with the distribution.
12 .\" 3. Neither the name of the University nor the names of its contributors
13 .\" may be used to endorse or promote products derived from this software
14 .\" without specific prior written permission.
16 .\" THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
17 .\" ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
18 .\" IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
19 .\" ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
20 .\" FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
21 .\" DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
22 .\" OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
23 .\" HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
24 .\" LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
25 .\" OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
28 .\" @(#)results.ms 6.2 (Berkeley) 4/16/91
34 The following tables indicate the results of our
36 Note that each table contains results for tests run
37 on two varieties of 4.2BSD file systems.
38 The first set of results is always for a file system
39 with a basic blocking factor of eight Kilobytes and a
40 fragment size of 1 Kilobyte. The second sets of measurements
41 are for file systems with a four Kilobyte block size and a
42 one Kilobyte fragment size.
43 The values in parenthesis indicate the percentage of CPU
44 time used by the test program.
45 In the case of the two disk arm tests,
46 the value in parenthesis indicates the sum of the percentage
47 of the test programs that were run.
48 Entries of ``n. m.'' indicate this value was not measured.
59 4.2BSD File Systems Tests - \fBVAX 11/750\fR
61 Logically Sequential Transfers
62 from an \fB8K/1K\fR 4.2BSD File System (Kbytes/sec.)
64 Test Emulex SC750/Eagle UDA50/RA81
66 1 Drive 2 Drives 1 Drive 2 Drives
68 read_8192 490 (69%) 620 (96%) 310 (44%) 520 (65%)
69 write_4096 380 (99%) 370 (99%) 370 (97%) 360 (98%)
70 write_8192 470 (99%) 470 (99%) 320 (71%) 410 (83%)
71 rewrite_8192 650 (99%) 620 (99%) 310 (50%) 450 (70%)
80 Logically Sequential Transfers
81 from \fB4K/1K\fR 4.2BSD File System (Kbytes/sec.)
83 Test Emulex SC750/Eagle UDA50/RA81
85 1 Drive 2 Drives 1 Drive 2 Drives
87 read_8192 300 (60%) 400 (84%) 210 (42%) 340 (77%)
88 write_4096 320 (98%) 320 (98%) 220 (67%) 290 (99%)
89 write_8192 340 (98%) 340 (99%) 220 (65%) 310 (98%)
90 rewrite_8192 450 (99%) 450 (98%) 230 (47%) 340 (78%)
94 Note that the rate of write operations on the VAX 11/750 are ultimately
95 CPU limited in some cases.
96 The write rates saturate the CPU at a lower bandwidth than the reads
97 because they must do disk allocation in addition to moving the data
98 from the user program to the disk.
99 The UDA50/RA81 saturates the CPU at a lower transfer rate for a given
100 operation than the SC750/Eagle because
101 it causes more memory contention with the CPU.
102 We do not know if this contention is caused by
103 the UNIBUS controller or the UDA50.
105 The following table reports the results of test runs on a VAX 11/780
106 with 4 Megabytes of main memory.
117 4.2BSD File Systems Tests - \fBVAX 11/780\fR
119 Logically Sequential Transfers
120 from an \fB8K/1K\fR 4.2BSD File System (Kbytes/sec.)
122 Test Emulex SC780/Eagle UDA50/RA81 Sys. Ind. 9900/Eagle
124 1 Drive 2 Drives 1 Drive 2 Drives 1 Drive 2 Drives
126 read_8192 560 (70%) 480 (58%) 360 (45%) 540 (72%) 340 (41%) 520 (66%)
127 write_4096 440 (98%) 440 (98%) 380 (99%) 480 (96%) 490 (96%) 440 (84%)
128 write_8192 490 (98%) 490 (98%) 220 (58%)* 480 (92%) 490 (80%) 430 (72%)
129 rewrite_8192 760 (100%) 560 (72%) 220 (50%)* 180 (52%)* 490 (60%) 520 (62%)
138 Logically Sequential Transfers
139 from an \fB4K/1K\fR 4.2BSD File System (Kbytes/sec.)
141 Test Emulex SC780/Eagle UDA50/RA81 Sys. Ind. 9900/Eagle
143 1 Drive 2 Drives 1 Drive 2 Drives 1 Drive 2 Drives
145 read_8192 490 (77%) 370 (66%) n.m. n.m. 200 (31%) 370 (56%)
146 write_4096 380 (98%) 370 (98%) n.m. n.m. 200 (46%) 370 (88%)
147 write_8192 380 (99%) 370 (97%) n.m. n.m. 200 (45%) 320 (76%)
148 rewrite_8192 490 (87%) 350 (66%) n.m. n.m. 200 (31%) 300 (46%)
150 * the operation of the hardware was suspect during these tests.
153 The dropoff in reading and writing rates for the two drive SC780/Eagle
154 tests are probably due to the file system using insufficient
155 rotational delay for these tests.
156 We have not fully investigated these times.
158 The following table compares data rates on VAX 11/750s directly
159 with those of VAX 11/780s using the UDA50/RA81 storage system.
170 4.2BSD File Systems Tests - \fBDEC UDA50 - 750 vs. 780\fR
172 Logically Sequential Transfers
173 from an \fB8K/1K\fR 4.2BSD File System (Kbytes/sec.)
175 Test VAX 11/750 UNIBUS VAX 11/780 UNIBUS
177 1 Drive 2 Drives 1 Drive 2 Drives
179 read_8192 310 (44%) 520 (84%) 360 (45%) 540 (72%)
180 write_4096 370 (97%) 360 (100%) 380 (99%) 480 (96%)
181 write_8192 320 (71%) 410 (96%) 220 (58%)* 480 (92%)
182 rewrite_8192 310 (50%) 450 (80%) 220 (50%)* 180 (52%)*
191 Logically Sequential Transfers
192 from an \fB4K/1K\fR 4.2BSD File System (Kbytes/sec.)
194 Test VAX 11/750 UNIBUS VAX 11/780 UNIBUS
196 1 Drive 2 Drives 1 Drive 2 Drives
198 read_8192 210 (42%) 342 (77%) n.m. n.m.
199 write_4096 215 (67%) 294 (99%) n.m. n.m.
200 write_8192 215 (65%) 305 (98%) n.m. n.m.
201 rewrite_8192 227 (47%) 336 (78%) n.m. n.m.
203 * the operation of the hardware was suspect during these tests.
206 The higher throughput available on VAX 11/780s is due to a number
208 The larger main memory size allows a larger file system cache.
209 The block allocation routines run faster, raising the upper limit
210 on the data rates in writing new files.
212 The next table makes the same comparison using an Emulex controller
224 4.2BSD File Systems Tests - \fBEmulex - 750 vs. 780\fR
226 Logically Sequential Transfers
227 from an \fB8K/1K\fR 4.2BSD File System (Kbytes/sec.)
229 Test VAX 11/750 CMI Bus VAX 11/780 SBI Bus
231 1 Drive 2 Drives 1 Drive 2 Drives
233 read_8192 490 (69%) 620 (96%) 560 (70%) 480 (58%)
234 write_4096 380 (99%) 370 (99%) 440 (98%) 440 (98%)
235 write_8192 470 (99%) 470 (99%) 490 (98%) 490 (98%)
236 rewrite_8192 650 (99%) 620 (99%) 760 (100%) 560 (72%)
245 Logically Sequential Transfers
246 from an \fB4K/1K\fR 4.2BSD File System (Kbytes/sec.)
248 Test VAX 11/750 CMI Bus VAX 11/780 SBI Bus
250 1 Drive 2 Drives 1 Drive 2 Drives
252 read_8192 300 (60%) 400 (84%) 490 (77%) 370 (66%)
253 write_4096 320 (98%) 320 (98%) 380 (98%) 370 (98%)
254 write_8192 340 (98%) 340 (99%) 380 (99%) 370 (97%)
255 rewrite_8192 450 (99%) 450 (98%) 490 (87%) 350 (66%)
259 The following table illustrates the evolution of our testing
260 process as both hardware and software problems effecting
261 the performance of the Emulex SC780 were corrected.
262 The software change was suggested to us by George Goble
263 of Purdue University.
265 The 4.2BSD handler for RH750/RH780 interfaced disk drives
266 contains several constants which to determine how
267 much time is provided between an interrupt signaling the completion
268 of a positioning command and the subsequent start of a data transfer
269 operation. These lead times are expressed as sectors of rotational delay.
270 If they are too small, an extra complete rotation will often be required
271 between a seek and subsequent read or write operation.
272 The higher bit rate and rotational speed of the 2351A Fujitsu
274 increasing these constants.
276 The hardware change involved allowing for slightly longer
277 delays in arbitrating for cycles on the SBI bus by
278 starting the bus arbitration cycle a little further ahead of
279 when the data was ready for transfer.
280 Finally we had to increase the rotational delay between consecutive
281 blocks in the file because
282 the higher bandwidth from the disk generated more memory contention,
283 which slowed down the processor.
295 4.2BSD File Systems Tests - \fBEmulex SC780 Disk Controller Evolution\fR
297 Logically Sequential Transfers
298 from an \fB8K/1K\fR 4.2BSD File System (Kbytes/sec.)
300 Test Inadequate Search Lead OK Search Lead OK Search Lead
301 Initial SBI Arbitration Init SBI Arb. Improved SBI Arb.
303 1 Drive 2 Drives 1 Drive 2 Drives 1 Drive 2 Drives
305 read_8192 320 370 440 (60%) n.m. 560 (70%) 480 (58%)
306 write_4096 250 270 300 (63%) n.m. 440 (98%) 440 (98%)
307 write_8192 250 280 340 (60%) n.m. 490 (98%) 490 (98%)
308 rewrite_8192 250 290 380 (48%) n.m. 760 (100%) 560 (72%)
318 Logically Sequential Transfers
319 from an \fB4K/1K\fR 4.2BSD File System (Kbytes/sec.)
321 Test Inadequate Search Lead OK Search Lead OK Search Lead
322 Initial SBI Arbitration Init SBI Arb. Improved SBI Arb.
324 1 Drive 2 Drives 1 Drive 2 Drives 1 Drive 2 Drives
326 read_8192 200 220 280 n.m. 490 (77%) 370 (66%)
327 write_4096 180 190 300 n.m. 380 (98%) 370 (98%)
328 write_8192 180 200 320 n.m. 380 (99%) 370 (97%)
329 rewrite_8192 190 200 340 n.m. 490 (87%) 350 (66%)