documentation/specification/expected_data.txt

XML via UDP - expected data from hosts
======================================


<<< Provisional, pjm2@kent.ac.uk >>>


Purpose of this document
------------------------

This document shall help to provide a separate party with
the knowledge required to use their own implementation of
a piece of host monitoring software.  In particular, this
document details the data that is to be expected from a
host.  This document is additional to
http://www.i-scream.org/cgi-bin/docs.cgi?doc=specification/xml_via_udp.txt


Background
----------

Hosts are expected to periodically send UDP packets to the
central monitoring system.  Such packets may contain various
pieces of information about the host, such as how much free
memory is remaining on the host, etc.  Some of this data is
deemed to be 'essential', some is 'expected' and the rest
is not necessary, but shall not be ignored by the central
monitoring system server.


Specification
-------------

The UDP packet must contain a complete and well-formed
piece of XML markup, describing the data that the host is
submitting.  In the event of a piece of data containing
an invalid character, such as a "<", it is the responsibilty
of the host to format this in a valid manner, for example
by replacing all occurrences of "<" with "&lt" before it is
sent in a UDP packet to the central monitoring system server.

Alternatively, the data may be wrapped up in special CDATA
tags.  For example, if we wish to send the following text:

    <>&!"

surrounded by <test_chars> tags then we can safely send this
by wrapping it up to produce the following XML snippet:

    <test_chars><![CDATA[<>&!"]]></test_chars>

The bare minimum that a host should send is the following: -

    <packet>
    </packet>

Note that every XML tag must also have a matching closing
tag.

However, such a packet is rather useless when it is
received by the server - it does not provide any useful
information.  As such, a packet will be rejected by the
server if it does not specify the machine_name, its
I.P. address (ip), a packet sequence number (seq_no)
and a date.  Note that the date is a long integer
representing the number of seconds since the epoch.
All of these are expected to be specified as an attribute
of the packet tag.  If they are not specified here, then
the packet will be rejected.

Thus, to ensure that a packet has a chance of being
acted upon, a host must specify at least four attribute
values - machine_name, ip, date and seq_no.  These must
be formed in the style shown below.

    <packet machine_name="raptor" ip="aaa.bbb.ccc.ddd"
            date="93456245245" seq_no="1234567">
    </packet>

Note that the server may run additional 'plugin' tools
that are designed to filter/reject packets under certain
conditions.  Please check with the server administrator
if you wish to know about the specific configuration of
your central monitoring system.

The I.P. address and machine_name could be found from
the UDP packet itself, but these remain essential so
as to provide resiliance from stray packets that do not
contain any valid or useful data.

Remember that malformed XML data would be rejected by the
central monitoring system immediately, without acting upon
it.


Classification of XML data in our system
----------------------------------------


ESSENTIAL: -

The following values are ESSENTIAL in each packet.  Any
packet without these will be rejected.  They MUST be
specified as attributes of the root node ("packet"): -

    ip
    seq_no
    date
    machine_name


EXPECTED: -

The following values are EXPECTED in each packet.  All
packets are RECOMMENDED to specify these values, as the
server is designed to permanently store such data in an
easy and quick to access manner.  It is not essential to
specify such values, but it does limit the usefulness of
the central monitoring system if these values are not
specified: -

    mac_address (attribute)
    os_name (attribute)
    os_version (attribute)
    cpus *
    load *
    memory *
    swap *
    disk *
    users (attribute)

* - These parameters are specified with tags of the same
    name.  They contain nested XML tags.  See later in
    this document for more details about that!

All of the values marked as being attributes above must be
specified as an attribute of the root tag ("packet")


OTHER: -

Any other arbitrary values may be specified by the host.
The only provision here is that they may not share the
same name as another tag or attribute within the same
heirarchical level.  The server will be able to handle
such clashes, however, it cannot be guaranteed which
of the duplicate parameters would gain priority.


Detail of a complete 'expected' XML packet via UDP
--------------------------------------------------

I shall start by showing an example of a complete
'expected' packet.  I shall then explain the contents
of it.

Here is an example of an XML packet, containing all of
the 'essential' and 'expected' data: -

<packet ip="" seq_no="" date="" machine_name=""
        mac_address="" os_name="" os_version=""
        user="">
    <cpus>
        <cpu0></cpu0>
        <cpu1></cpu1>
        <cpu2></cpu2>
    </cpus>
    <load>
        <load1><load1>
        <load5><load5>
        <load15><load15>
    </load>
    <memory>
        <free></free>
        <total></total>
    </memory>
    <swap>
        <free></free>
        <total></total>
    </swap>
    <disk>
        <p0>
            <free></free>
            <total></total>
            <mounted></mounted>
            <label></label>
        </p0>
        <p1>
            <free></free>
            <total></total>
            <mount></mount>
            <label></label>
        </p1>
    </disk>
</packet>

Note that linefeeds and spaces would normally not be present.
These are only shown above for structure clarity.

Note that all of the essential attributes have been specified
as attributes of the root node.  The expected attribute
values are also attributes of the root node.

The <cpus> tag begins a list of cpu load information.
Each subtag herein must be of the form cpux, where x is
the number of the processor in multiprocessor systems.
Note that this numbering starts from zero.

The <load> tag begins a list of system load information.
The load tag is only expected to contain 3 subtags, named
"load1", "load5" and "load15".  These represent the average
load 1 minute ago, 5 minutes ago and 15 minutes ago
respectively.

The <memory> tag begins a list of system memory information.
This tag is only expected to contain two subtags, as shown
in the example above.  These two subtags specify how much
memory the system has free and how much memory the system
has in total.

The <swap> tag is identical in nature to the <memory> tag.

The <disk> tag begins a list of partition information.
Each subtag must be of the form px, where x is the number
of the disk partition that is unique for that particular
system.  Note that the numbering of px is expected to
start from zero.  Each <px> tag contains the following
subtags: -
    <free> - the free partition space in bytes.
    <total> - the total partition size in bytes.
    <mount> - the mount point, eg, "/var".
    <label> - the label of the physical disk.

The <disk_total> tag begins a list of total disk space,
similar in implementation to the <disk_free> tag.

Any other values and attributes may be specified in the
packet, providing they are not specified in the same
heirarchical level as another attribute or value with
the same name.


Flexibility and Minimising bandwidth
------------------------------------

The following issues still stand: -

The means of submitting host data via UDP containing XML
markup is provided so that future customisation is easily
possible.  It would be possible to easily tailor a custom
piece of host monitoring software to provide exactly what
data is desired for adequate monitoring.

Some of the XML markup demonstrated above contains a lot
of rendundant features.  For example, it is not necessary
to lay the contents out neatly (although this certainly
helps visualise the contents).

The amount of data sent within each UDP packet may be (in
some cases, vastly) reduced by using some of the ideas
described below: -

  - Remove unnecessary linefeeds and 'white space'

  - If a single piece of data is to be represented, it
    will usually occupy less space if it is stored as
    an attribute to a tag, rather than within a pair
    of tags.

  - Comments within the XML may be useful for testing
    purposes, however, the server ignores all comments
    so these can be removed to reduce packet sizes.

Taking the above into account, here is an example of some
ideally formed XML to be contained within a UDP packet: -

    <packet machine_name="raptor" ip="aaa.bbb.ccc.ddd"
    ><freespace><drive1>23677</drive1><drive2>23534</d
    rive2><drive3>10390</drive3></freespace></packet>

Notice how all unnecessary 'white space' and linefeeds have
been removed.  The comment has also been removed.  Values
such as "machine_name" and "ip" have both been stored as an
attribute of the root node ("packet") as this results in
a smaller packet size.


About
-----

This document was written by Paul Mutton [pjm2@kent.ac.uk]
for use by the team working on a 3rd year Computer Science 
project called "i-scream". More details can be found on the 
project website - http://www.i-scream.org

 
Revision:	1.7
Committed:	Sun Aug 1 10:39:55 2004 UTC (21 years, 3 months ago) by tdb
Content type:	text/plain
Branch:	MAIN
CVS Tags:	HEAD
Changes since 1.6:	+4 -4 lines
Error occurred while calculating annotation data.
Log Message:	Catch a lot of old URL's and update them. Also remove a couple of old files that aren't used.
#	Content
1	XML via UDP - expected data from hosts
2	======================================
3
4
5	<<< Provisional, pjm2@kent.ac.uk >>>
6
7
8	Purpose of this document
9	------------------------
10
11	This document shall help to provide a separate party with
12	the knowledge required to use their own implementation of
13	a piece of host monitoring software. In particular, this
14	document details the data that is to be expected from a
15	host. This document is additional to
16	http://www.i-scream.org/cgi-bin/docs.cgi?doc=specification/xml_via_udp.txt
17
18
19	Background
20	----------
21
22	Hosts are expected to periodically send UDP packets to the
23	central monitoring system. Such packets may contain various
24	pieces of information about the host, such as how much free
25	memory is remaining on the host, etc. Some of this data is
26	deemed to be 'essential', some is 'expected' and the rest
27	is not necessary, but shall not be ignored by the central
28	monitoring system server.
29
30
31	Specification
32	-------------
33
34	The UDP packet must contain a complete and well-formed
35	piece of XML markup, describing the data that the host is
36	submitting. In the event of a piece of data containing
37	an invalid character, such as a "<", it is the responsibilty
38	of the host to format this in a valid manner, for example
39	by replacing all occurrences of "<" with "&lt" before it is
40	sent in a UDP packet to the central monitoring system server.
41
42	Alternatively, the data may be wrapped up in special CDATA
43	tags. For example, if we wish to send the following text:
44
45	<>&!"
46
47	surrounded by <test_chars> tags then we can safely send this
48	by wrapping it up to produce the following XML snippet:
49
50	<test_chars><![CDATA[<>&!"]]></test_chars>
51
52	The bare minimum that a host should send is the following: -
53
54	<packet>
55	</packet>
56
57	Note that every XML tag must also have a matching closing
58	tag.
59
60	However, such a packet is rather useless when it is
61	received by the server - it does not provide any useful
62	information. As such, a packet will be rejected by the
63	server if it does not specify the machine_name, its
64	I.P. address (ip), a packet sequence number (seq_no)
65	and a date. Note that the date is a long integer
66	representing the number of seconds since the epoch.
67	All of these are expected to be specified as an attribute
68	of the packet tag. If they are not specified here, then
69	the packet will be rejected.
70
71	Thus, to ensure that a packet has a chance of being
72	acted upon, a host must specify at least four attribute
73	values - machine_name, ip, date and seq_no. These must
74	be formed in the style shown below.
75
76	<packet machine_name="raptor" ip="aaa.bbb.ccc.ddd"
77	date="93456245245" seq_no="1234567">
78	</packet>
79
80	Note that the server may run additional 'plugin' tools
81	that are designed to filter/reject packets under certain
82	conditions. Please check with the server administrator
83	if you wish to know about the specific configuration of
84	your central monitoring system.
85
86	The I.P. address and machine_name could be found from
87	the UDP packet itself, but these remain essential so
88	as to provide resiliance from stray packets that do not
89	contain any valid or useful data.
90
91	Remember that malformed XML data would be rejected by the
92	central monitoring system immediately, without acting upon
93	it.
94
95
96	Classification of XML data in our system
97	----------------------------------------
98
99
100	ESSENTIAL: -
101
102	The following values are ESSENTIAL in each packet. Any
103	packet without these will be rejected. They MUST be
104	specified as attributes of the root node ("packet"): -
105
106	ip
107	seq_no
108	date
109	machine_name
110
111
112	EXPECTED: -
113
114	The following values are EXPECTED in each packet. All
115	packets are RECOMMENDED to specify these values, as the
116	server is designed to permanently store such data in an
117	easy and quick to access manner. It is not essential to
118	specify such values, but it does limit the usefulness of
119	the central monitoring system if these values are not
120	specified: -
121
122	mac_address (attribute)
123	os_name (attribute)
124	os_version (attribute)
125	cpus *
126	load *
127	memory *
128	swap *
129	disk *
130	users (attribute)
131
132	* - These parameters are specified with tags of the same
133	name. They contain nested XML tags. See later in
134	this document for more details about that!
135
136	All of the values marked as being attributes above must be
137	specified as an attribute of the root tag ("packet")
138
139
140	OTHER: -
141
142	Any other arbitrary values may be specified by the host.
143	The only provision here is that they may not share the
144	same name as another tag or attribute within the same
145	heirarchical level. The server will be able to handle
146	such clashes, however, it cannot be guaranteed which
147	of the duplicate parameters would gain priority.
148
149
150	Detail of a complete 'expected' XML packet via UDP
151	--------------------------------------------------
152
153	I shall start by showing an example of a complete
154	'expected' packet. I shall then explain the contents
155	of it.
156
157	Here is an example of an XML packet, containing all of
158	the 'essential' and 'expected' data: -
159
160	<packet ip="" seq_no="" date="" machine_name=""
161	mac_address="" os_name="" os_version=""
162	user="">
163	<cpus>
164	<cpu0></cpu0>
165	<cpu1></cpu1>
166	<cpu2></cpu2>
167	</cpus>
168	<load>
169	<load1><load1>
170	<load5><load5>
171	<load15><load15>
172	</load>
173	<memory>
174	<free></free>
175	<total></total>
176	</memory>
177	<swap>
178	<free></free>
179	<total></total>
180	</swap>
181	<disk>
182	<p0>
183	<free></free>
184	<total></total>
185	<mounted></mounted>
186	<label></label>
187	</p0>
188	<p1>
189	<free></free>
190	<total></total>
191	<mount></mount>
192	<label></label>
193	</p1>
194	</disk>
195	</packet>
196
197	Note that linefeeds and spaces would normally not be present.
198	These are only shown above for structure clarity.
199
200	Note that all of the essential attributes have been specified
201	as attributes of the root node. The expected attribute
202	values are also attributes of the root node.
203
204	The <cpus> tag begins a list of cpu load information.
205	Each subtag herein must be of the form cpux, where x is
206	the number of the processor in multiprocessor systems.
207	Note that this numbering starts from zero.
208
209	The <load> tag begins a list of system load information.
210	The load tag is only expected to contain 3 subtags, named
211	"load1", "load5" and "load15". These represent the average
212	load 1 minute ago, 5 minutes ago and 15 minutes ago
213	respectively.
214
215	The <memory> tag begins a list of system memory information.
216	This tag is only expected to contain two subtags, as shown
217	in the example above. These two subtags specify how much
218	memory the system has free and how much memory the system
219	has in total.
220
221	The <swap> tag is identical in nature to the <memory> tag.
222
223	The <disk> tag begins a list of partition information.
224	Each subtag must be of the form px, where x is the number
225	of the disk partition that is unique for that particular
226	system. Note that the numbering of px is expected to
227	start from zero. Each <px> tag contains the following
228	subtags: -
229	<free> - the free partition space in bytes.
230	<total> - the total partition size in bytes.
231	<mount> - the mount point, eg, "/var".
232	<label> - the label of the physical disk.
233
234	The <disk_total> tag begins a list of total disk space,
235	similar in implementation to the <disk_free> tag.
236
237	Any other values and attributes may be specified in the
238	packet, providing they are not specified in the same
239	heirarchical level as another attribute or value with
240	the same name.
241
242
243	Flexibility and Minimising bandwidth
244	------------------------------------
245
246	The following issues still stand: -
247
248	The means of submitting host data via UDP containing XML
249	markup is provided so that future customisation is easily
250	possible. It would be possible to easily tailor a custom
251	piece of host monitoring software to provide exactly what
252	data is desired for adequate monitoring.
253
254	Some of the XML markup demonstrated above contains a lot
255	of rendundant features. For example, it is not necessary
256	to lay the contents out neatly (although this certainly
257	helps visualise the contents).
258
259	The amount of data sent within each UDP packet may be (in
260	some cases, vastly) reduced by using some of the ideas
261	described below: -
262
263	- Remove unnecessary linefeeds and 'white space'
264
265	- If a single piece of data is to be represented, it
266	will usually occupy less space if it is stored as
267	an attribute to a tag, rather than within a pair
268	of tags.
269
270	- Comments within the XML may be useful for testing
271	purposes, however, the server ignores all comments
272	so these can be removed to reduce packet sizes.
273
274	Taking the above into account, here is an example of some
275	ideally formed XML to be contained within a UDP packet: -
276
277	<packet machine_name="raptor" ip="aaa.bbb.ccc.ddd"
278	><freespace><drive1>23677</drive1><drive2>23534</d
279	rive2><drive3>10390</drive3></freespace></packet>
280
281	Notice how all unnecessary 'white space' and linefeeds have
282	been removed. The comment has also been removed. Values
283	such as "machine_name" and "ip" have both been stored as an
284	attribute of the root node ("packet") as this results in
285	a smaller packet size.
286
287
288	About
289	-----
290
291	This document was written by Paul Mutton [pjm2@kent.ac.uk]
292	for use by the team working on a 3rd year Computer Science
293	project called "i-scream". More details can be found on the
294	project website - http://www.i-scream.org
295
296