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. All advertising materials mentioning features or use of this software
13 .\" must display the following acknowledgement:
14 .\" This product includes software developed by the University of
15 .\" California, Berkeley and its contributors.
16 .\" 4. Neither the name of the University nor the names of its contributors
17 .\" may be used to endorse or promote products derived from this software
18 .\" without specific prior written permission.
20 .\" THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
21 .\" ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22 .\" IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23 .\" ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
24 .\" FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25 .\" DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26 .\" OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27 .\" HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28 .\" LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29 .\" OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
32 .\" @(#)results.ms 6.2 (Berkeley) 4/16/91
38 The following tables indicate the results of our
40 Note that each table contains results for tests run
41 on two varieties of 4.2BSD file systems.
42 The first set of results is always for a file system
43 with a basic blocking factor of eight Kilobytes and a
44 fragment size of 1 Kilobyte. The second sets of measurements
45 are for file systems with a four Kilobyte block size and a
46 one Kilobyte fragment size.
47 The values in parenthesis indicate the percentage of CPU
48 time used by the test program.
49 In the case of the two disk arm tests,
50 the value in parenthesis indicates the sum of the percentage
51 of the test programs that were run.
52 Entries of ``n. m.'' indicate this value was not measured.
63 4.2BSD File Systems Tests - \fBVAX 11/750\fR
65 Logically Sequential Transfers
66 from an \fB8K/1K\fR 4.2BSD File System (Kbytes/sec.)
68 Test Emulex SC750/Eagle UDA50/RA81
70 1 Drive 2 Drives 1 Drive 2 Drives
72 read_8192 490 (69%) 620 (96%) 310 (44%) 520 (65%)
73 write_4096 380 (99%) 370 (99%) 370 (97%) 360 (98%)
74 write_8192 470 (99%) 470 (99%) 320 (71%) 410 (83%)
75 rewrite_8192 650 (99%) 620 (99%) 310 (50%) 450 (70%)
84 Logically Sequential Transfers
85 from \fB4K/1K\fR 4.2BSD File System (Kbytes/sec.)
87 Test Emulex SC750/Eagle UDA50/RA81
89 1 Drive 2 Drives 1 Drive 2 Drives
91 read_8192 300 (60%) 400 (84%) 210 (42%) 340 (77%)
92 write_4096 320 (98%) 320 (98%) 220 (67%) 290 (99%)
93 write_8192 340 (98%) 340 (99%) 220 (65%) 310 (98%)
94 rewrite_8192 450 (99%) 450 (98%) 230 (47%) 340 (78%)
98 Note that the rate of write operations on the VAX 11/750 are ultimately
99 CPU limited in some cases.
100 The write rates saturate the CPU at a lower bandwidth than the reads
101 because they must do disk allocation in addition to moving the data
102 from the user program to the disk.
103 The UDA50/RA81 saturates the CPU at a lower transfer rate for a given
104 operation than the SC750/Eagle because
105 it causes more memory contention with the CPU.
106 We do not know if this contention is caused by
107 the UNIBUS controller or the UDA50.
109 The following table reports the results of test runs on a VAX 11/780
110 with 4 Megabytes of main memory.
121 4.2BSD File Systems Tests - \fBVAX 11/780\fR
123 Logically Sequential Transfers
124 from an \fB8K/1K\fR 4.2BSD File System (Kbytes/sec.)
126 Test Emulex SC780/Eagle UDA50/RA81 Sys. Ind. 9900/Eagle
128 1 Drive 2 Drives 1 Drive 2 Drives 1 Drive 2 Drives
130 read_8192 560 (70%) 480 (58%) 360 (45%) 540 (72%) 340 (41%) 520 (66%)
131 write_4096 440 (98%) 440 (98%) 380 (99%) 480 (96%) 490 (96%) 440 (84%)
132 write_8192 490 (98%) 490 (98%) 220 (58%)* 480 (92%) 490 (80%) 430 (72%)
133 rewrite_8192 760 (100%) 560 (72%) 220 (50%)* 180 (52%)* 490 (60%) 520 (62%)
142 Logically Sequential Transfers
143 from an \fB4K/1K\fR 4.2BSD File System (Kbytes/sec.)
145 Test Emulex SC780/Eagle UDA50/RA81 Sys. Ind. 9900/Eagle
147 1 Drive 2 Drives 1 Drive 2 Drives 1 Drive 2 Drives
149 read_8192 490 (77%) 370 (66%) n.m. n.m. 200 (31%) 370 (56%)
150 write_4096 380 (98%) 370 (98%) n.m. n.m. 200 (46%) 370 (88%)
151 write_8192 380 (99%) 370 (97%) n.m. n.m. 200 (45%) 320 (76%)
152 rewrite_8192 490 (87%) 350 (66%) n.m. n.m. 200 (31%) 300 (46%)
154 * the operation of the hardware was suspect during these tests.
157 The dropoff in reading and writing rates for the two drive SC780/Eagle
158 tests are probably due to the file system using insufficient
159 rotational delay for these tests.
160 We have not fully investigated these times.
162 The following table compares data rates on VAX 11/750s directly
163 with those of VAX 11/780s using the UDA50/RA81 storage system.
174 4.2BSD File Systems Tests - \fBDEC UDA50 - 750 vs. 780\fR
176 Logically Sequential Transfers
177 from an \fB8K/1K\fR 4.2BSD File System (Kbytes/sec.)
179 Test VAX 11/750 UNIBUS VAX 11/780 UNIBUS
181 1 Drive 2 Drives 1 Drive 2 Drives
183 read_8192 310 (44%) 520 (84%) 360 (45%) 540 (72%)
184 write_4096 370 (97%) 360 (100%) 380 (99%) 480 (96%)
185 write_8192 320 (71%) 410 (96%) 220 (58%)* 480 (92%)
186 rewrite_8192 310 (50%) 450 (80%) 220 (50%)* 180 (52%)*
195 Logically Sequential Transfers
196 from an \fB4K/1K\fR 4.2BSD File System (Kbytes/sec.)
198 Test VAX 11/750 UNIBUS VAX 11/780 UNIBUS
200 1 Drive 2 Drives 1 Drive 2 Drives
202 read_8192 210 (42%) 342 (77%) n.m. n.m.
203 write_4096 215 (67%) 294 (99%) n.m. n.m.
204 write_8192 215 (65%) 305 (98%) n.m. n.m.
205 rewrite_8192 227 (47%) 336 (78%) n.m. n.m.
207 * the operation of the hardware was suspect during these tests.
210 The higher throughput available on VAX 11/780s is due to a number
212 The larger main memory size allows a larger file system cache.
213 The block allocation routines run faster, raising the upper limit
214 on the data rates in writing new files.
216 The next table makes the same comparison using an Emulex controller
228 4.2BSD File Systems Tests - \fBEmulex - 750 vs. 780\fR
230 Logically Sequential Transfers
231 from an \fB8K/1K\fR 4.2BSD File System (Kbytes/sec.)
233 Test VAX 11/750 CMI Bus VAX 11/780 SBI Bus
235 1 Drive 2 Drives 1 Drive 2 Drives
237 read_8192 490 (69%) 620 (96%) 560 (70%) 480 (58%)
238 write_4096 380 (99%) 370 (99%) 440 (98%) 440 (98%)
239 write_8192 470 (99%) 470 (99%) 490 (98%) 490 (98%)
240 rewrite_8192 650 (99%) 620 (99%) 760 (100%) 560 (72%)
249 Logically Sequential Transfers
250 from an \fB4K/1K\fR 4.2BSD File System (Kbytes/sec.)
252 Test VAX 11/750 CMI Bus VAX 11/780 SBI Bus
254 1 Drive 2 Drives 1 Drive 2 Drives
256 read_8192 300 (60%) 400 (84%) 490 (77%) 370 (66%)
257 write_4096 320 (98%) 320 (98%) 380 (98%) 370 (98%)
258 write_8192 340 (98%) 340 (99%) 380 (99%) 370 (97%)
259 rewrite_8192 450 (99%) 450 (98%) 490 (87%) 350 (66%)
263 The following table illustrates the evolution of our testing
264 process as both hardware and software problems effecting
265 the performance of the Emulex SC780 were corrected.
266 The software change was suggested to us by George Goble
267 of Purdue University.
269 The 4.2BSD handler for RH750/RH780 interfaced disk drives
270 contains several constants which to determine how
271 much time is provided between an interrupt signaling the completion
272 of a positioning command and the subsequent start of a data transfer
273 operation. These lead times are expressed as sectors of rotational delay.
274 If they are too small, an extra complete rotation will often be required
275 between a seek and subsequent read or write operation.
276 The higher bit rate and rotational speed of the 2351A Fujitsu
278 increasing these constants.
280 The hardware change involved allowing for slightly longer
281 delays in arbitrating for cycles on the SBI bus by
282 starting the bus arbitration cycle a little further ahead of
283 when the data was ready for transfer.
284 Finally we had to increase the rotational delay between consecutive
285 blocks in the file because
286 the higher bandwidth from the disk generated more memory contention,
287 which slowed down the processor.
299 4.2BSD File Systems Tests - \fBEmulex SC780 Disk Controller Evolution\fR
301 Logically Sequential Transfers
302 from an \fB8K/1K\fR 4.2BSD File System (Kbytes/sec.)
304 Test Inadequate Search Lead OK Search Lead OK Search Lead
305 Initial SBI Arbitration Init SBI Arb. Improved SBI Arb.
307 1 Drive 2 Drives 1 Drive 2 Drives 1 Drive 2 Drives
309 read_8192 320 370 440 (60%) n.m. 560 (70%) 480 (58%)
310 write_4096 250 270 300 (63%) n.m. 440 (98%) 440 (98%)
311 write_8192 250 280 340 (60%) n.m. 490 (98%) 490 (98%)
312 rewrite_8192 250 290 380 (48%) n.m. 760 (100%) 560 (72%)
322 Logically Sequential Transfers
323 from an \fB4K/1K\fR 4.2BSD File System (Kbytes/sec.)
325 Test Inadequate Search Lead OK Search Lead OK Search Lead
326 Initial SBI Arbitration Init SBI Arb. Improved SBI Arb.
328 1 Drive 2 Drives 1 Drive 2 Drives 1 Drive 2 Drives
330 read_8192 200 220 280 n.m. 490 (77%) 370 (66%)
331 write_4096 180 190 300 n.m. 380 (98%) 370 (98%)
332 write_8192 180 200 320 n.m. 380 (99%) 370 (97%)
333 rewrite_8192 190 200 340 n.m. 490 (87%) 350 (66%)