1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
|
<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 3.2 Draft//EN">
<HTML>
<HEAD>
<TITLE>SQUID Frequently Asked Questions: DISKD</TITLE>
</HEAD>
<BODY>
<A HREF="FAQ-23.html">Next</A>
<A HREF="FAQ-21.html">Previous</A>
<A HREF="FAQ.html#toc22">Contents</A>
<HR>
<H2><A NAME="s22">22. DISKD</A></H2>
<H2><A NAME="ss22.1">22.1 What is DISKD?</A>
</H2>
<P>DISKD refers to some features in Squid-2.4 to improve Disk I/O performance.
The basic idea is that each <EM>cache_dir</EM> has its own <EM>diskd</EM> child process.
The diskd process performs all disk I/O operations (open, close, read, write, unlink)
for the cache_dir. Message queues are used to send requests and responses between
the Squid and diskd processes. Shared memory is used for chunks of data to
be read and written.
<P>
<H2><A NAME="ss22.2">22.2 Does it perform better?</A>
</H2>
<P>Yes. We benchmarked Squid-2.4 with DISKD at the
<A HREF="http://polygraph.ircache.net/Results/bakeoff-2/">Second IRCache Bake-Off</A>.
The results are also described
<A HREF="/Benchmarking/bakeoff-02/">here</A>.
At the bakeoff, we got 160 req/sec with diskd. Without diskd, we'd have gotten about 40 req/sec.
<P>
<H2><A NAME="ss22.3">22.3 What do I need to use it?</A>
</H2>
<P>
<OL>
<LI>Squid-2.4</LI>
<LI>Your operating system must support message queues.</LI>
<LI>Your operating system must support shared memory.</LI>
</OL>
<P>
<H2><A NAME="ss22.4">22.4 If I use DISKD, do I have to wipe out my current cache?</A>
</H2>
<P>No. Diskd uses the same storage scheme as the standard "UFS"
type. It only changes how I/O is performed.
<P>
<H2><A NAME="ss22.5">22.5 How do I configure message queues?</A>
</H2>
<P>Most Unix operating systems have message queue support
by default. One way to check is to see if you have
an <EM>ipcs</EM> command.
<P>
<P>However, you will likely need to increase the message
queue parameters for Squid. Message queue implementations
normally have the following parameters:
<DL>
<DT><B>MSGMNB</B><DD><P>Maximum number of bytes in a single queue.
<DT><B>MSGMNI</B><DD><P>Maximum number of message queue identifiers.
<DT><B>MSGSEG</B><DD><P>Maximum number of message segments.
<DT><B>MSGMAX</B><DD><P>Maximum size of a message segment.
<DT><B>MSGTQL</B><DD><P>Maximum number of messages in the whole system.
</DL>
<P>
<P>The messages between Squid and diskd are 32 bytes. Thus, MSGMAX
should be 32 or greater. You may want to set it to a larger
value, just to be safe.
<P>
<P>We'll have two queues for each <EM>cache_dir</EM> -- one in each direction.
So, MSGMNI needs to be at least two times the number of <EM>cache_dir</EM>'s.
<P>
<P>MSGMNB and MSGTQL affect how many messages can be in the queues
at one time. I've found that 75 messages per queue is about
the limit of decent performance. Thus, MSGMNB must be
at least 75*MSGMAX, and MSGTQL must be at least 75 times
the number of <EM>cache_dir</EM>'s.
<P>
<H3>FreeBSD</H3>
<P>Your kernel must have
<PRE>
options SYSVMSG
</PRE>
<P>
<P>You can set the parameters in the kernel as follows. This is just
an example. Make sure the values are appropriate for your system:
<PRE>
options MSGMNB=16384 # max # of bytes in a queue
options MSGMNI=41 # number of message queue identifiers
options MSGSEG=2049 # number of message segments
options MSGSSZ=64 # size of a message segment
options MSGTQL=512 # max messages in system
</PRE>
<P>
<H3>Digital Unix</H3>
<P>Message queue support seems to be in the kernel
by default. Setting the options is as follows:
<PRE>
options MSGMNB="8192" # max # bytes on queue
options MSGMNI="31" # # of message queue identifiers
options MSGMAX="2049" # max message size
options MSGTQL="1024" # # of system message headers
</PRE>
<P>
<P>by
<A HREF="mailto:B.C.Phillips at massey dot ac dot nz">Brenden Phillips</A><P>If you have a newer version (DU64), then you can probably use
<EM>sysconfig</EM> instead. To see what the current IPC settings are run
<PRE>
# sysconfig -q ipc
</PRE>
To change them make a file like this called ipc.stanza:
<PRE>
ipc:
msg-max = 2049
msg-mni = 31
msg-tql = 1024
msg-mnb = 8192
</PRE>
then run
<PRE>
# sysconfigdb -a -f ipc.stanza
</PRE>
You have to reboot for the change to take effect.
<P>
<P>
<P>
<H3>Linux</H3>
<P>In my limited browsing on Linux, I didn't see any way to change
message queue parameters except to modify the include files
and build a new kernel. On my system, the file
is <EM>/usr/src/linux/include/linux/msg.h</EM>.
<P>
<H3>Solaris</H3>
<P>Refer to
<A HREF="http://www.sunworld.com/sunworldonline/swol-11-1997/swol-11-insidesolaris.html">Demangling Message Queues</A> in Sunworld Magazine.
<P>
<P>I don't think the above article really tells you how to set the parameters.
You do it in <EM>/etc/system</EM> with lines like this:
<PRE>
set msgsys:msginfo_msgmax=2049
set msgsys:msginfo_msgmnb=8192
set msgsys:msginfo_msgmni=31
set msgsys:msginfo_msgssz=64
set msgsys:msginfo_msgtql=1024
</PRE>
<P>Of course, you must reboot whenever you modify <EM>/etc/system</EM>
before changes take effect.
<P>
<H2><A NAME="ss22.6">22.6 How do I configure shared memory?</A>
</H2>
<P>Shared memory uses a set of parameters similar to the ones for message
queues. The Squid DISKD implementation uses one shared memory area
for each cache_dir. Each shared memory area is about
800 kilobytes in size. You may need to modify your system's
shared memory parameters:
<P>
<P>
<DL>
<DT><B>SHMSEG</B><DD><P>Maximum number of shared memory segments per process.
<DT><B>SHMMNI</B><DD><P>Maximum number of shared memory segments for the whole system.
<DT><B>SHMMAX</B><DD><P>Largest shared memory segment size allowed.
<DT><B>SHMALL</B><DD><P>Total amount of shared memory that can be used.
</DL>
<P>
<P>For Squid and DISKD, <EM>SHMMNI</EM> and <EM>SHMMNI</EM> must be greater than
or equal to the number of <EM>cache_dir</EM>'s that you have. <EM>SHMMAX</EM>
must be at least 800 kilobytes. <EM>SHMALL</EM> must be at least
<EM>SHMMAX</EM> 800 kilobytes multiplied by the number of <EM>cache_dir</EM>'s.
<P>
<H3>FreeBSD</H3>
<P>Your kernel must have
<PRE>
options SYSVSHM
</PRE>
<P>
<P>You can set the parameters in the kernel as follows. This is just
an example. Make sure the values are appropriate for your system:
<PRE>
options SHMSEG=16 # max shared mem id's per process
options SHMMNI=32 # max shared mem id's per system
options SHMMAX=2097152 # max shared memory segment size (bytes)
options SHMALL=4096 # max amount of shared memory (pages)
</PRE>
<P>
<H3>Digital Unix</H3>
<P>Message queue support seems to be in the kernel
by default. Setting the options is as follows:
<PRE>
options SHMSEG="16" # max shared mem id's per process
options SHMMNI="32" # max shared mem id's per system
options SHMMAX="2097152" # max shared memory segment size (bytes)
options SHMALL=4096 # max amount of shared memory (pages)
</PRE>
<P>
<P>by
<A HREF="mailto:B.C.Phillips at massey dot ac dot nz">Brenden Phillips</A><P>If you have a newer version (DU64), then you can probably use
<EM>sysconfig</EM> instead. To see what the current IPC settings are run
<PRE>
# sysconfig -q ipc
</PRE>
To change them make a file like this called ipc.stanza:
<PRE>
ipc:
shm-seg = 16
shm-mni = 32
shm-max = 2097152
shm-all = 4096
</PRE>
then run
<PRE>
# sysconfigdb -a -f ipc.stanza
</PRE>
You have to reboot for the change to take effect.
<P>
<P>
<H3>Linux</H3>
<P>In my limited browsing on Linux, I didn't see any way to change
shared memory parameters except to modify the include files
and build a new kernel. On my system, the file
is <EM>/usr/src/linux/include/asm-i386/shmparam.h</EM>
<P>
<P>Oh, it looks like you can change <EM>SHMMAX</EM> by writing
the file <EM>/proc/sys/kernel/shmmax</EM>.
<P>
<H3>Solaris</H3>
<P>Refer to
<A HREF="http://www.sunworld.com/swol-09-1997/swol-09-insidesolaris.html">Shared memory uncovered</A>
in Sunworld Magazine.
<P>
<P>To set the values, you can put these lines in <EM>/etc/system</EM>:
<PRE>
set shmsys:shminfo_shmmax=2097152
set shmsys:shminfo_shmmni=32
set shmsys:shminfo_shmseg=16
</PRE>
<P>
<H2><A NAME="ss22.7">22.7 Sometimes shared memory and message queues aren't released when Squid exits.</A>
</H2>
<P>Yes, this is a little problem sometimes. Seems like the operating system
gets confused and doesn't always release shared memory and message
queue resources when processes exit, especially if they exit abnormally.
To fix it you can ``manually'' clear the resources with the <EM>ipcs</EM> command.
Add this command into your <EM>RunCache</EM> or <EM>squid_start</EM>
script:
<PRE>
ipcs | grep '^[mq]' | awk '{printf "ipcrm -%s %s\n", $1, $2}' | /bin/sh
</PRE>
<P>
<H2><A NAME="ss22.8">22.8 What are the Q1 and Q2 parameters?</A>
</H2>
<P>In the source code, these are called <EM>magic1</EM> and <EM>magic2</EM>.
These numbers refer to the number of oustanding requests on a message
queue. They are specified on the <EM>cache_dir</EM> option line, after
the L1 and L2 directories:
<PRE>
cache_dir diskd -1 /cache1 1024 16 256 64 72
</PRE>
<P>If there are more than Q1 messages outstanding, then the main Squid
process ``blocks'' for a little bit until the diskd process services
some of the messages and sends back some replies.
<P>If there are more than Q2 messages outstanding, then Squid will
intentionally fail to open disk files for reading and writing.
This is a load-shedding mechanism. If your cache gets really really
busy and the disks can not keep up, Squid bypasses the disks until
the load goes down again.
<P>Reasonable values for Q1 and Q2 are 64 and 72, respectively.
<P>
<P>
<HR>
<A HREF="FAQ-23.html">Next</A>
<A HREF="FAQ-21.html">Previous</A>
<A HREF="FAQ.html#toc22">Contents</A>
</BODY>
</HTML>
|
