Code

Imported upstream version 1.3rc9.
[pkg-rrdtool.git] / doc / rrdcreate.html
1 <?xml version="1.0" ?>
2 <!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Strict//EN" "http://www.w3.org/TR/xhtml1/DTD/xhtml1-strict.dtd">
3 <html xmlns="http://www.w3.org/1999/xhtml">
4 <head>
5 <title>rrdcreate</title>
6 <meta http-equiv="content-type" content="text/html; charset=utf-8" />
7 <link rev="made" href="mailto:root@localhost" />
8 </head>
10 <body style="background-color: white">
12 <p><a name="__index__"></a></p>
13 <!-- INDEX BEGIN -->
14 <!--
16 <ul>
18         <li><a href="#name">NAME</a></li>
19         <li><a href="#synopsis">SYNOPSIS</a></li>
20         <li><a href="#description">DESCRIPTION</a></li>
21         <li><a href="#aberrant_behavior_detection_with_holtwinters_forecasting">Aberrant Behavior Detection with Holt-Winters Forecasting</a></li>
22         <li><a href="#the_heartbeat_and_the_step">The HEARTBEAT and the STEP</a></li>
23         <li><a href="#how_to_measure">HOW TO MEASURE</a></li>
24         <li><a href="#example">EXAMPLE</a></li>
25         <li><a href="#example_2">EXAMPLE 2</a></li>
26         <li><a href="#example_3">EXAMPLE 3</a></li>
27         <li><a href="#author">AUTHOR</a></li>
28 </ul>
29 -->
30 <!-- INDEX END -->
32 <p>
33 </p>
34 <h1><a name="name">NAME</a></h1>
35 <p>rrdcreate - Set up a new Round Robin Database</p>
36 <p>
37 </p>
38 <hr />
39 <h1><a name="synopsis">SYNOPSIS</a></h1>
40 <p><strong>rrdtool</strong> <strong>create</strong> <em>filename</em>
41 [<strong>--start</strong>|<strong>-b</strong>&nbsp;<em>start&nbsp;time</em>]
42 [<strong>--step</strong>|<strong>-s</strong>&nbsp;<em>step</em>]
43 [<strong>DS:</strong><em>ds-name</em><strong>:</strong><em>DST</em><strong>:</strong><em>dst&nbsp;arguments</em>]
44 [<strong>RRA:</strong><em>CF</em><strong>:</strong><em>cf&nbsp;arguments</em>]</p>
45 <p>
46 </p>
47 <hr />
48 <h1><a name="description">DESCRIPTION</a></h1>
49 <p>The create function of RRDtool lets you set up new Round Robin
50 Database (<strong>RRD</strong>) files.  The file is created at its final, full size
51 and filled with <em>*UNKNOWN*</em> data.</p>
52 <dl>
53 <dt><strong><a name="item_filename"><em>filename</em></a></strong></dt>
55 <dd>
56 <p>The name of the <strong>RRD</strong> you want to create. <strong>RRD</strong> files should end
57 with the extension <em>.rrd</em>. However, <strong>RRDtool</strong> will accept any
58 filename.</p>
59 </dd>
60 <dt><strong><a name="item_time"><strong>--start</strong>|<strong>-b</strong> <em>start time</em> (default: now - 10s)</a></strong></dt>
62 <dd>
63 <p>Specifies the time in seconds since 1970-01-01 UTC when the first
64 value should be added to the <strong>RRD</strong>. <strong>RRDtool</strong> will not accept
65 any data timed before or at the time specified.</p>
66 <p>See also AT-STYLE TIME SPECIFICATION section in the
67 <em>rrdfetch</em> documentation for other ways to specify time.</p>
68 </dd>
69 <dt><strong><a name="item_step"><strong>--step</strong>|<strong>-s</strong> <em>step</em> (default: 300 seconds)</a></strong></dt>
71 <dd>
72 <p>Specifies the base interval in seconds with which data will be fed
73 into the <strong>RRD</strong>.</p>
74 </dd>
75 <dt><strong><a name="item_ds_3ads_2dname_3adst_3adst_arguments"><strong>DS:</strong><em>ds-name</em><strong>:</strong><em>DST</em><strong>:</strong><em>dst arguments</em></a></strong></dt>
77 <dd>
78 <p>A single <strong>RRD</strong> can accept input from several data sources (<strong>DS</strong>),
79 for example incoming and outgoing traffic on a specific communication
80 line. With the <strong>DS</strong> configuration option you must define some basic
81 properties of each data source you want to store in the <strong>RRD</strong>.</p>
82 <p><em>ds-name</em> is the name you will use to reference this particular data
83 source from an <strong>RRD</strong>. A <em>ds-name</em> must be 1 to 19 characters long in
84 the characters [a-zA-Z0-9_].</p>
85 <p><em>DST</em> defines the Data Source Type. The remaining arguments of a
86 data source entry depend on the data source type. For GAUGE, COUNTER,
87 DERIVE, and ABSOLUTE the format for a data source entry is:</p>
88 <p><strong>DS:</strong><em>ds-name</em><strong>:</strong><em>GAUGE | COUNTER | DERIVE | ABSOLUTE</em><strong>:</strong><em>heartbeat</em><strong>:</strong><em>min</em><strong>:</strong><em>max</em></p>
89 <p>For COMPUTE data sources, the format is:</p>
90 <p><strong>DS:</strong><em>ds-name</em><strong>:</strong><em>COMPUTE</em><strong>:</strong><em>rpn-expression</em></p>
91 <p>In order to decide which data source type to use, review the
92 definitions that follow. Also consult the section on ``HOW TO MEASURE''
93 for further insight.</p>
94 <dl>
95 <dt><strong><a name="item_gauge"><strong>GAUGE</strong></a></strong></dt>
97 <dd>
98 <p>is for things like temperatures or number of people in a room or the
99 value of a RedHat share.</p>
100 </dd>
101 <dt><strong><a name="item_counter"><strong>COUNTER</strong></a></strong></dt>
103 <dd>
104 <p>is for continuous incrementing counters like the ifInOctets counter in
105 a router. The <strong>COUNTER</strong> data source assumes that the counter never
106 decreases, except when a counter overflows.  The update function takes
107 the overflow into account.  The counter is stored as a per-second
108 rate. When the counter overflows, RRDtool checks if the overflow
109 happened at the 32bit or 64bit border and acts accordingly by adding
110 an appropriate value to the result.</p>
111 </dd>
112 <dt><strong><a name="item_derive"><strong>DERIVE</strong></a></strong></dt>
114 <dd>
115 <p>will store the derivative of the line going from the last to the
116 current value of the data source. This can be useful for gauges, for
117 example, to measure the rate of people entering or leaving a
118 room. Internally, derive works exactly like COUNTER but without
119 overflow checks. So if your counter does not reset at 32 or 64 bit you
120 might want to use DERIVE and combine it with a MIN value of 0.</p>
121 <dl>
122 <dt><strong><a name="item_note_on_counter_vs_derive">NOTE on COUNTER vs DERIVE</a></strong></dt>
124 <dd>
125 <p>by Don Baarda &lt;<a href="mailto:don.baarda@baesystems.com">don.baarda@baesystems.com</a>&gt;</p>
126 <p>If you cannot tolerate ever mistaking the occasional counter reset for a
127 legitimate counter wrap, and would prefer ``Unknowns'' for all legitimate
128 counter wraps and resets, always use DERIVE with min=0. Otherwise, using
129 COUNTER with a suitable max will return correct values for all legitimate
130 counter wraps, mark some counter resets as ``Unknown'', but can mistake some
131 counter resets for a legitimate counter wrap.</p>
132 <p>For a 5 minute step and 32-bit counter, the probability of mistaking a
133 counter reset for a legitimate wrap is arguably about 0.8% per 1Mbps of
134 maximum bandwidth. Note that this equates to 80% for 100Mbps interfaces, so
135 for high bandwidth interfaces and a 32bit counter, DERIVE with min=0 is
136 probably preferable. If you are using a 64bit counter, just about any max
137 setting will eliminate the possibility of mistaking a reset for a counter
138 wrap.</p>
139 </dd>
140 </dl>
141 </dd>
142 <dt><strong><a name="item_absolute"><strong>ABSOLUTE</strong></a></strong></dt>
144 <dd>
145 <p>is for counters which get reset upon reading. This is used for fast counters
146 which tend to overflow. So instead of reading them normally you reset them
147 after every read to make sure you have a maximum time available before the
148 next overflow. Another usage is for things you count like number of messages
149 since the last update.</p>
150 </dd>
151 <dt><strong><a name="item_compute"><strong>COMPUTE</strong></a></strong></dt>
153 <dd>
154 <p>is for storing the result of a formula applied to other data sources
155 in the <strong>RRD</strong>. This data source is not supplied a value on update, but
156 rather its Primary Data Points (PDPs) are computed from the PDPs of
157 the data sources according to the rpn-expression that defines the
158 formula. Consolidation functions are then applied normally to the PDPs
159 of the COMPUTE data source (that is the rpn-expression is only applied
160 to generate PDPs). In database software, such data sets are referred
161 to as ``virtual'' or ``computed'' columns.</p>
162 </dd>
163 </dl>
164 <p><em>heartbeat</em> defines the maximum number of seconds that may pass
165 between two updates of this data source before the value of the
166 data source is assumed to be <em>*UNKNOWN*</em>.</p>
167 <p><em>min</em> and <em>max</em> define the expected range values for data supplied by a
168 data source. If <em>min</em> and/or <em>max</em> any value outside the defined range
169 will be regarded as <em>*UNKNOWN*</em>. If you do not know or care about min and
170 max, set them to U for unknown. Note that min and max always refer to the
171 processed values of the DS. For a traffic-<strong>COUNTER</strong> type DS this would be
172 the maximum and minimum data-rate expected from the device.</p>
173 <p><em>If information on minimal/maximal expected values is available,
174 always set the min and/or max properties. This will help RRDtool in
175 doing a simple sanity check on the data supplied when running update.</em></p>
176 <p><em>rpn-expression</em> defines the formula used to compute the PDPs of a
177 COMPUTE data source from other data sources in the same &lt;RRD&gt;. It is
178 similar to defining a <strong>CDEF</strong> argument for the graph command. Please
179 refer to that manual page for a list and description of RPN operations
180 supported. For COMPUTE data sources, the following RPN operations are
181 not supported: COUNT, PREV, TIME, and LTIME. In addition, in defining
182 the RPN expression, the COMPUTE data source may only refer to the
183 names of data source listed previously in the create command. This is
184 similar to the restriction that <strong>CDEF</strong>s must refer only to <strong>DEF</strong>s
185 and <strong>CDEF</strong>s previously defined in the same graph command.</p>
186 </dd>
187 <dt><strong><a name="item_rra_3acf_3acf_arguments"><strong>RRA:</strong><em>CF</em><strong>:</strong><em>cf arguments</em></a></strong></dt>
189 <dd>
190 <p>The purpose of an <strong>RRD</strong> is to store data in the round robin archives
191 (<strong>RRA</strong>). An archive consists of a number of data values or statistics for
192 each of the defined data-sources (<strong>DS</strong>) and is defined with an <strong>RRA</strong> line.</p>
193 <p>When data is entered into an <strong>RRD</strong>, it is first fit into time slots
194 of the length defined with the <strong>-s</strong> option, thus becoming a <em>primary
195 data point</em>.</p>
196 <p>The data is also processed with the consolidation function (<em>CF</em>) of
197 the archive. There are several consolidation functions that
198 consolidate primary data points via an aggregate function: <strong>AVERAGE</strong>,
199 <strong>MIN</strong>, <strong>MAX</strong>, <strong>LAST</strong>.</p>
200 <dl>
201 <dt><strong><a name="item_average">AVERAGE</a></strong></dt>
203 <dd>
204 <p>the average of the data points is stored.</p>
205 </dd>
206 <dt><strong><a name="item_min">MIN</a></strong></dt>
208 <dd>
209 <p>the smallest of the data points is stored.</p>
210 </dd>
211 <dt><strong><a name="item_max">MAX</a></strong></dt>
213 <dd>
214 <p>the largest of the data points is stored.</p>
215 </dd>
216 <dt><strong><a name="item_last">LAST</a></strong></dt>
218 <dd>
219 <p>the last data points is used.</p>
220 </dd>
221 </dl>
222 <p>Note that data aggregation inevitably leads to loss of precision and
223 information. The trick is to pick the aggregate function such that the
224 <em>interesting</em> properties of your data is kept across the aggregation
225 process.</p>
226 <p>The format of <strong>RRA</strong> line for these
227 consolidation functions is:</p>
228 <p><strong>RRA:</strong><em>AVERAGE | MIN | MAX | LAST</em><strong>:</strong><em>xff</em><strong>:</strong><em>steps</em><strong>:</strong><em>rows</em></p>
229 <p><em>xff</em> The xfiles factor defines what part of a consolidation interval may
230 be made up from <em>*UNKNOWN*</em> data while the consolidated value is still
231 regarded as known. It is given as the ratio of allowed <em>*UNKNOWN*</em> PDPs
232 to the number of PDPs in the interval. Thus, it ranges from 0 to 1 (exclusive).</p>
233 <p><em>steps</em> defines how many of these <em>primary data points</em> are used to build
234 a <em>consolidated data point</em> which then goes into the archive.</p>
235 <p><em>rows</em> defines how many generations of data values are kept in an <strong>RRA</strong>.</p>
236 </dd>
237 </dl>
238 <p>
239 </p>
240 <hr />
241 <h1><a name="aberrant_behavior_detection_with_holtwinters_forecasting">Aberrant Behavior Detection with Holt-Winters Forecasting</a></h1>
242 <p>In addition to the aggregate functions, there are a set of specialized
243 functions that enable <strong>RRDtool</strong> to provide data smoothing (via the
244 Holt-Winters forecasting algorithm), confidence bands, and the
245 flagging aberrant behavior in the data source time series:</p>
246 <ul>
247 <li>
248 <p><strong>RRA:</strong><em>HWPREDICT</em><strong>:</strong><em>rows</em><strong>:</strong><em>alpha</em><strong>:</strong><em>beta</em><strong>:</strong><em>seasonal period</em>[<strong>:</strong><em>rra-num</em>]</p>
249 </li>
250 <li>
251 <p><strong>RRA:</strong><em>MHWPREDICT</em><strong>:</strong><em>rows</em><strong>:</strong><em>alpha</em><strong>:</strong><em>beta</em><strong>:</strong><em>seasonal period</em>[<strong>:</strong><em>rra-num</em>]</p>
252 </li>
253 <li>
254 <p><strong>RRA:</strong><em>SEASONAL</em><strong>:</strong><em>seasonal period</em><strong>:</strong><em>gamma</em><strong>:</strong><em>rra-num</em>[<strong>:smoothing-window=</strong><em>fraction</em>]</p>
255 </li>
256 <li>
257 <p><strong>RRA:</strong><em>DEVSEASONAL</em><strong>:</strong><em>seasonal period</em><strong>:</strong><em>gamma</em><strong>:</strong><em>rra-num</em>[<strong>:smoothing-window=</strong><em>fraction</em>]</p>
258 </li>
259 <li>
260 <p><strong>RRA:</strong><em>DEVPREDICT</em><strong>:</strong><em>rows</em><strong>:</strong><em>rra-num</em></p>
261 </li>
262 <li>
263 <p><strong>RRA:</strong><em>FAILURES</em><strong>:</strong><em>rows</em><strong>:</strong><em>threshold</em><strong>:</strong><em>window length</em><strong>:</strong><em>rra-num</em></p>
264 </li>
265 </ul>
266 <p>These <strong>RRAs</strong> differ from the true consolidation functions in several ways.
267 First, each of the <strong>RRA</strong>s is updated once for every primary data point.
268 Second, these <strong>RRAs</strong> are interdependent. To generate real-time confidence
269 bounds, a matched set of SEASONAL, DEVSEASONAL, DEVPREDICT, and either
270 HWPREDICT or MHWPREDICT must exist. Generating smoothed values of the primary
271 data points requires a SEASONAL <strong>RRA</strong> and either an HWPREDICT or MHWPREDICT 
272 <strong>RRA</strong>. Aberrant behavior detection requires FAILURES, DEVSEASONAL, SEASONAL,
273 and either HWPREDICT or MHWPREDICT.</p>
274 <p>The predicted, or smoothed, values are stored in the HWPREDICT or MHWPREDICT
275 <strong>RRA</strong>. HWPREDICT and MHWPREDICT are actually two variations on the
276 Holt-Winters method. They are interchangeable. Both attempt to decompose data
277 into three components: a baseline, a trend, and a seasonal coefficient.
278 HWPREDICT adds its seasonal coefficient to the baseline to form a prediction, whereas
279 MHWPREDICT multiplies its seasonal coefficient by the baseline to form a
280 prediction. The difference is noticeable when the baseline changes
281 significantly in the course of a season; HWPREDICT will predict the seasonality
282 to stay constant as the baseline changes, but MHWPREDICT will predict the
283 seasonality to grow or shrink in proportion to the baseline. The proper choice
284 of method depends on the thing being modeled. For simplicity, the rest of this
285 discussion will refer to HWPREDICT, but MHWPREDICT may be substituted in its
286 place.</p>
287 <p>The predicted deviations are stored in DEVPREDICT (think a standard deviation
288 which can be scaled to yield a confidence band). The FAILURES <strong>RRA</strong> stores 
289 binary indicators. A 1 marks the indexed observation as failure; that is, the 
290 number of confidence bounds violations in the preceding window of observations 
291 met or exceeded a specified threshold. An example of using these <strong>RRAs</strong> to graph 
292 confidence bounds and failures appears in <a href="././rrdgraph.html">the rrdgraph manpage</a>.</p>
293 <p>The SEASONAL and DEVSEASONAL <strong>RRAs</strong> store the seasonal coefficients for the
294 Holt-Winters forecasting algorithm and the seasonal deviations, respectively.
295 There is one entry per observation time point in the seasonal cycle. For
296 example, if primary data points are generated every five minutes and the
297 seasonal cycle is 1 day, both SEASONAL and DEVSEASONAL will have 288 rows.</p>
298 <p>In order to simplify the creation for the novice user, in addition to
299 supporting explicit creation of the HWPREDICT, SEASONAL, DEVPREDICT,
300 DEVSEASONAL, and FAILURES <strong>RRAs</strong>, the <strong>RRDtool</strong> create command supports
301 implicit creation of the other four when HWPREDICT is specified alone and
302 the final argument <em>rra-num</em> is omitted.</p>
303 <p><em>rows</em> specifies the length of the <strong>RRA</strong> prior to wrap around. Remember
304 that there is a one-to-one correspondence between primary data points and
305 entries in these RRAs. For the HWPREDICT CF, <em>rows</em> should be larger than
306 the <em>seasonal period</em>. If the DEVPREDICT <strong>RRA</strong> is implicitly created, the
307 default number of rows is the same as the HWPREDICT <em>rows</em> argument. If the
308 FAILURES <strong>RRA</strong> is implicitly created, <em>rows</em> will be set to the <em>seasonal
309 period</em> argument of the HWPREDICT <strong>RRA</strong>. Of course, the <strong>RRDtool</strong>
310 <em>resize</em> command is available if these defaults are not sufficient and the
311 creator wishes to avoid explicit creations of the other specialized function
312 <strong>RRAs</strong>.</p>
313 <p><em>seasonal period</em> specifies the number of primary data points in a seasonal
314 cycle. If SEASONAL and DEVSEASONAL are implicitly created, this argument for
315 those <strong>RRAs</strong> is set automatically to the value specified by HWPREDICT. If
316 they are explicitly created, the creator should verify that all three
317 <em>seasonal period</em> arguments agree.</p>
318 <p><em>alpha</em> is the adaption parameter of the intercept (or baseline)
319 coefficient in the Holt-Winters forecasting algorithm. See <a href="././rrdtool.html">the rrdtool manpage</a> for a
320 description of this algorithm. <em>alpha</em> must lie between 0 and 1. A value
321 closer to 1 means that more recent observations carry greater weight in
322 predicting the baseline component of the forecast. A value closer to 0 means
323 that past history carries greater weight in predicting the baseline
324 component.</p>
325 <p><em>beta</em> is the adaption parameter of the slope (or linear trend) coefficient
326 in the Holt-Winters forecasting algorithm. <em>beta</em> must lie between 0 and 1
327 and plays the same role as <em>alpha</em> with respect to the predicted linear
328 trend.</p>
329 <p><em>gamma</em> is the adaption parameter of the seasonal coefficients in the
330 Holt-Winters forecasting algorithm (HWPREDICT) or the adaption parameter in
331 the exponential smoothing update of the seasonal deviations. It must lie
332 between 0 and 1. If the SEASONAL and DEVSEASONAL <strong>RRAs</strong> are created
333 implicitly, they will both have the same value for <em>gamma</em>: the value
334 specified for the HWPREDICT <em>alpha</em> argument. Note that because there is
335 one seasonal coefficient (or deviation) for each time point during the
336 seasonal cycle, the adaptation rate is much slower than the baseline. Each
337 seasonal coefficient is only updated (or adapts) when the observed value
338 occurs at the offset in the seasonal cycle corresponding to that
339 coefficient.</p>
340 <p>If SEASONAL and DEVSEASONAL <strong>RRAs</strong> are created explicitly, <em>gamma</em> need not
341 be the same for both. Note that <em>gamma</em> can also be changed via the
342 <strong>RRDtool</strong> <em>tune</em> command.</p>
343 <p><em>smoothing-window</em> specifies the fraction of a season that should be
344 averaged around each point. By default, the value of <em>smoothing-window</em> is
345 0.05, which means each value in SEASONAL and DEVSEASONAL will be occasionally
346 replaced by averaging it with its (<em>seasonal period</em>*0.05) nearest neighbors.
347 Setting <em>smoothing-window</em> to zero will disable the running-average smoother
348 altogether.</p>
349 <p><em>rra-num</em> provides the links between related <strong>RRAs</strong>. If HWPREDICT is
350 specified alone and the other <strong>RRAs</strong> are created implicitly, then
351 there is no need to worry about this argument. If <strong>RRAs</strong> are created
352 explicitly, then carefully pay attention to this argument. For each
353 <strong>RRA</strong> which includes this argument, there is a dependency between
354 that <strong>RRA</strong> and another <strong>RRA</strong>. The <em>rra-num</em> argument is the 1-based
355 index in the order of <strong>RRA</strong> creation (that is, the order they appear
356 in the <em>create</em> command). The dependent <strong>RRA</strong> for each <strong>RRA</strong>
357 requiring the <em>rra-num</em> argument is listed here:</p>
358 <ul>
359 <li>
360 <p>HWPREDICT <em>rra-num</em> is the index of the SEASONAL <strong>RRA</strong>.</p>
361 </li>
362 <li>
363 <p>SEASONAL <em>rra-num</em> is the index of the HWPREDICT <strong>RRA</strong>.</p>
364 </li>
365 <li>
366 <p>DEVPREDICT <em>rra-num</em> is the index of the DEVSEASONAL <strong>RRA</strong>.</p>
367 </li>
368 <li>
369 <p>DEVSEASONAL <em>rra-num</em> is the index of the HWPREDICT <strong>RRA</strong>.</p>
370 </li>
371 <li>
372 <p>FAILURES <em>rra-num</em> is the index of the DEVSEASONAL <strong>RRA</strong>.</p>
373 </li>
374 </ul>
375 <p><em>threshold</em> is the minimum number of violations (observed values outside
376 the confidence bounds) within a window that constitutes a failure. If the
377 FAILURES <strong>RRA</strong> is implicitly created, the default value is 7.</p>
378 <p><em>window length</em> is the number of time points in the window. Specify an
379 integer greater than or equal to the threshold and less than or equal to 28.
380 The time interval this window represents depends on the interval between
381 primary data points. If the FAILURES <strong>RRA</strong> is implicitly created, the
382 default value is 9.</p>
383 <p>
384 </p>
385 <hr />
386 <h1><a name="the_heartbeat_and_the_step">The HEARTBEAT and the STEP</a></h1>
387 <p>Here is an explanation by Don Baarda on the inner workings of RRDtool.
388 It may help you to sort out why all this *UNKNOWN* data is popping
389 up in your databases:</p>
390 <p>RRDtool gets fed samples/updates at arbitrary times. From these it builds Primary
391 Data Points (PDPs) on every ``step'' interval. The PDPs are
392 then accumulated into the RRAs.</p>
393 <p>The ``heartbeat'' defines the maximum acceptable interval between
394 samples/updates. If the interval between samples is less than ``heartbeat'',
395 then an average rate is calculated and applied for that interval. If
396 the interval between samples is longer than ``heartbeat'', then that
397 entire interval is considered ``unknown''. Note that there are other
398 things that can make a sample interval ``unknown'', such as the rate
399 exceeding limits, or a sample that was explicitly marked as unknown.</p>
400 <p>The known rates during a PDP's ``step'' interval are used to calculate
401 an average rate for that PDP. If the total ``unknown'' time accounts for
402 more than <strong>half</strong> the ``step'', the entire PDP is marked
403 as ``unknown''. This means that a mixture of known and ``unknown'' sample
404 times in a single PDP ``step'' may or may not add up to enough ``known''
405 time to warrent for a known PDP.</p>
406 <p>The ``heartbeat'' can be short (unusual) or long (typical) relative to
407 the ``step'' interval between PDPs. A short ``heartbeat'' means you
408 require multiple samples per PDP, and if you don't get them mark the
409 PDP unknown. A long heartbeat can span multiple ``steps'', which means
410 it is acceptable to have multiple PDPs calculated from a single
411 sample. An extreme example of this might be a ``step'' of 5 minutes and a
412 ``heartbeat'' of one day, in which case a single sample every day will
413 result in all the PDPs for that entire day period being set to the
414 same average rate. <em>-- Don Baarda &lt;<a href="mailto:don.baarda@baesystems.com">don.baarda@baesystems.com</a>&gt;</em></p>
415 <pre>
416        time|
417        axis|
418  begin__|00|
419         |01|
420        u|02|----* sample1, restart &quot;hb&quot;-timer
421        u|03|   /
422        u|04|  /
423        u|05| /
424        u|06|/     &quot;hbt&quot; expired
425        u|07|
426         |08|----* sample2, restart &quot;hb&quot; 
427         |09|   / 
428         |10|  /
429        u|11|----* sample3, restart &quot;hb&quot;
430        u|12|   /
431        u|13|  /
432  step1_u|14| /
433        u|15|/     &quot;swt&quot; expired
434        u|16|
435         |17|----* sample4, restart &quot;hb&quot;, create &quot;pdp&quot; for step1 = 
436         |18|   /  = unknown due to 10 &quot;u&quot; labled secs &gt; 0.5 * step
437         |19|  /
438         |20| /
439         |21|----* sample5, restart &quot;hb&quot;
440         |22|   /
441         |23|  /
442         |24|----* sample6, restart &quot;hb&quot;
443         |25|   /
444         |26|  /
445         |27|----* sample7, restart &quot;hb&quot;
446  step2__|28|   /
447         |22|  /
448         |23|----* sample8, restart &quot;hb&quot;, create &quot;pdp&quot; for step1, create &quot;cdp&quot; 
449         |24|   /
450         |25|  /</pre>
451 <p>graphics by <em><a href="mailto:vladimir.lavrov@desy.de">vladimir.lavrov@desy.de</a></em>.</p>
452 <p>
453 </p>
454 <hr />
455 <h1><a name="how_to_measure">HOW TO MEASURE</a></h1>
456 <p>Here are a few hints on how to measure:</p>
457 <dl>
458 <dt><strong><a name="item_temperature">Temperature</a></strong></dt>
460 <dd>
461 <p>Usually you have some type of meter you can read to get the temperature.
462 The temperature is not really connected with a time. The only connection is
463 that the temperature reading happened at a certain time. You can use the
464 <strong>GAUGE</strong> data source type for this. RRDtool will then record your reading
465 together with the time.</p>
466 </dd>
467 <dt><strong><a name="item_mail_messages">Mail Messages</a></strong></dt>
469 <dd>
470 <p>Assume you have a method to count the number of messages transported by
471 your mailserver in a certain amount of time, giving you data like '5
472 messages in the last 65 seconds'. If you look at the count of 5 like an
473 <strong>ABSOLUTE</strong> data type you can simply update the RRD with the number 5 and the
474 end time of your monitoring period. RRDtool will then record the number of
475 messages per second. If at some later stage you want to know the number of
476 messages transported in a day, you can get the average messages per second
477 from RRDtool for the day in question and multiply this number with the
478 number of seconds in a day. Because all math is run with Doubles, the
479 precision should be acceptable.</p>
480 </dd>
481 <dt><strong><a name="item_it_27s_always_a_rate">It's always a Rate</a></strong></dt>
483 <dd>
484 <p>RRDtool stores rates in amount/second for COUNTER, DERIVE and ABSOLUTE
485 data.  When you plot the data, you will get on the y axis
486 amount/second which you might be tempted to convert to an absolute
487 amount by multiplying by the delta-time between the points. RRDtool
488 plots continuous data, and as such is not appropriate for plotting
489 absolute amounts as for example ``total bytes'' sent and received in a
490 router. What you probably want is plot rates that you can scale to
491 bytes/hour, for example, or plot absolute amounts with another tool
492 that draws bar-plots, where the delta-time is clear on the plot for
493 each point (such that when you read the graph you see for example GB
494 on the y axis, days on the x axis and one bar for each day).</p>
495 </dd>
496 </dl>
497 <p>
498 </p>
499 <hr />
500 <h1><a name="example">EXAMPLE</a></h1>
501 <pre>
502  rrdtool create temperature.rrd --step 300 \
503   DS:temp:GAUGE:600:-273:5000 \
504   RRA:AVERAGE:0.5:1:1200 \
505   RRA:MIN:0.5:12:2400 \
506   RRA:MAX:0.5:12:2400 \
507   RRA:AVERAGE:0.5:12:2400</pre>
508 <p>This sets up an <strong>RRD</strong> called <em>temperature.rrd</em> which accepts one
509 temperature value every 300 seconds. If no new data is supplied for
510 more than 600 seconds, the temperature becomes <em>*UNKNOWN*</em>.  The
511 minimum acceptable value is -273 and the maximum is 5'000.</p>
512 <p>A few archive areas are also defined. The first stores the
513 temperatures supplied for 100 hours (1'200 * 300 seconds = 100
514 hours). The second RRA stores the minimum temperature recorded over
515 every hour (12 * 300 seconds = 1 hour), for 100 days (2'400 hours). The
516 third and the fourth RRA's do the same for the maximum and
517 average temperature, respectively.</p>
518 <p>
519 </p>
520 <hr />
521 <h1><a name="example_2">EXAMPLE 2</a></h1>
522 <pre>
523  rrdtool create monitor.rrd --step 300        \
524    DS:ifOutOctets:COUNTER:1800:0:4294967295   \
525    RRA:AVERAGE:0.5:1:2016                     \
526    RRA:HWPREDICT:1440:0.1:0.0035:288</pre>
527 <p>This example is a monitor of a router interface. The first <strong>RRA</strong> tracks the
528 traffic flow in octets; the second <strong>RRA</strong> generates the specialized
529 functions <strong>RRAs</strong> for aberrant behavior detection. Note that the <em>rra-num</em>
530 argument of HWPREDICT is missing, so the other <strong>RRAs</strong> will implicitly be
531 created with default parameter values. In this example, the forecasting
532 algorithm baseline adapts quickly; in fact the most recent one hour of
533 observations (each at 5 minute intervals) accounts for 75% of the baseline
534 prediction. The linear trend forecast adapts much more slowly. Observations
535 made during the last day (at 288 observations per day) account for only
536 65% of the predicted linear trend. Note: these computations rely on an
537 exponential smoothing formula described in the LISA 2000 paper.</p>
538 <p>The seasonal cycle is one day (288 data points at 300 second intervals), and
539 the seasonal adaption parameter will be set to 0.1. The RRD file will store 5
540 days (1'440 data points) of forecasts and deviation predictions before wrap
541 around. The file will store 1 day (a seasonal cycle) of 0-1 indicators in
542 the FAILURES <strong>RRA</strong>.</p>
543 <p>The same RRD file and <strong>RRAs</strong> are created with the following command,
544 which explicitly creates all specialized function <strong>RRAs</strong>.</p>
545 <pre>
546  rrdtool create monitor.rrd --step 300 \
547    DS:ifOutOctets:COUNTER:1800:0:4294967295 \
548    RRA:AVERAGE:0.5:1:2016 \
549    RRA:HWPREDICT:1440:0.1:0.0035:288:3 \
550    RRA:SEASONAL:288:0.1:2 \
551    RRA:DEVPREDICT:1440:5 \
552    RRA:DEVSEASONAL:288:0.1:2 \
553    RRA:FAILURES:288:7:9:5</pre>
554 <p>Of course, explicit creation need not replicate implicit create, a
555 number of arguments could be changed.</p>
556 <p>
557 </p>
558 <hr />
559 <h1><a name="example_3">EXAMPLE 3</a></h1>
560 <pre>
561  rrdtool create proxy.rrd --step 300 \
562    DS:Total:DERIVE:1800:0:U  \
563    DS:Duration:DERIVE:1800:0:U  \
564    DS:AvgReqDur:COMPUTE:Duration,Requests,0,EQ,1,Requests,IF,/ \
565    RRA:AVERAGE:0.5:1:2016</pre>
566 <p>This example is monitoring the average request duration during each 300 sec
567 interval for requests processed by a web proxy during the interval.
568 In this case, the proxy exposes two counters, the number of requests
569 processed since boot and the total cumulative duration of all processed
570 requests. Clearly these counters both have some rollover point, but using the
571 DERIVE data source also handles the reset that occurs when the web proxy is
572 stopped and restarted.</p>
573 <p>In the <strong>RRD</strong>, the first data source stores the requests per second rate
574 during the interval. The second data source stores the total duration of all
575 requests processed during the interval divided by 300. The COMPUTE data source
576 divides each PDP of the AccumDuration by the corresponding PDP of
577 TotalRequests and stores the average request duration. The remainder of the
578 RPN expression handles the divide by zero case.</p>
579 <p>
580 </p>
581 <hr />
582 <h1><a name="author">AUTHOR</a></h1>
583 <p>Tobias Oetiker &lt;<a href="mailto:tobi@oetiker.ch">tobi@oetiker.ch</a>&gt;</p>
585 </body>
587 </html>