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
|
<!--startcut ==========================================================-->
<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 3.2//EN">
<HTML>
<HEAD>
<title>Addendum to the Slashdot Effect Internet Paper LG #38</title>
</HEAD>
<BODY BGCOLOR="#FFFFFF" TEXT="#000000" LINK="#0000FF" VLINK="#0000AF"
ALINK="#FF0000">
<!--endcut ============================================================-->
<H4>
"Linux Gazette...<I>making Linux just a little more fun!</I>"
</H4>
<P> <HR> <P>
<!--===================================================================-->
<center>
<H1><font color="maroon">Addendum to The Slashdot Effect</font></H1>
<H4>By <a href="http://ssadler.phy.bnl.gov/adler/adler">Stephen Adler</a></H4>
</center>
<P>
<hr width="80%">
<center>Addendum to
<a href="./adler1.html">
The Slashdot Effect Internet Paper</a>
<p>
</center>
<blockquote>
After the Slashdot Effect paper was announced on /., a very large
surge in hits was experienced by the server hosting the
paper. Interest in the original paper detailing the /. effect due to
other articles hosted by this server, was greater by a factor of 4
than any of the previous articles. As of Feb 15, 1999, about two weeks
after the announcement, the paper has been viewed by over 17,000
people. This is compared to the other three articles mentioned in the
/. paper of which the Stallman article, which was read by the greatest
number of people, was viewed by just over 4000 people. It was
announced on /. over 3 weeks ago. What follows are the httpd server
logs during the time that the /. effect paper was announced on
Slashdot. In short, this addendum is presenting the Slashdot effect on
the Slashdot effect paper.
</blockquote>
<p>
<hr width="80%">
<p>
The apache httpd server logs were scanned once more, to measure the hit
rate in raw hits per minute, two weeks after the /. effect paper was
announced on Slashdot. The /. effect on this paper was quite
remarkable and an addendum to the original /. effect paper was
due. The surge in hits/minute was phenomenal as attested by the
following plots. The first and main plot, in Figure 1, is that of the
raw hits/minute shown over a period of 1 week. The red line marks the
time of the /. announcement of the /. effect paper. The blue line
indicates the time when the paper was announced on Linux Weekly
News. The purple lines indicate time marks on 24 hour intervals, with
the first one marking Feb 1st 00:00 hours. The paper was announced on
Feb 1st, at 9:40pm EST. The /. effect took place about 15 minutes
later. (See Figures 2 and 3.) The highest hit rate was measured right
up front. It was literally a wave front hitting the server. The hit
rate quieted down over night, with a very pronounced lull starting about
3am for about 2 hours. One would postulate that the /. server was down
during that period. The rate picked up again and climbed to a second
high, not as high as the initial surge, at about 9:30am. This rate
held for about another 5 hours and began to fall rather drastically
around 3pm EST the day after the announcement on Slashdot. The hit
rate persistence is consistent with the announcement staying on the
front page of Slashdot. The announcement left the front page of /. at
about the time the hit rate fell on the server.
<p>
<center>
<table>
<tr>
<td>
<a href="./gx/adler/SDEPlot1.jpg"><img src="./gx/adler/SDEPlot1Small.jpg"></a>
<caption align="bottom"><font size=-1>
<center>Figure 1</center>
Httpd apache log files showing the number of hits per minute received
by the server spanning a period of 1 week. The red line indicates the
time which Slashdot announced the paper.
</font>
</table>
</center>
<p>
What follows is a table of plots of the same data but on a finer time
scale. (9 min/bin as opposed to 50 min/bin in the plot in Figure 1.)
The purpose is for one to better study the data. The purple lines are
now time marker on 6 hour intervals. The limits on the time axis
overlap between plots to help the reader follow the continuity of the
data. Because of binning effects, the overlap regions of the data from
one plot to the next look slightly different, and this dissimilarity
should be ignored.
<p>
The first item to note is the fact that the highest hit rate occurs
just after the announcement on Slashdot. The second item to note is
the resurgence in hit rate the following day, while the announcement
is still on the front page. This is an indication that the readership
of Slashdot is covered 24 hours a day, or around the world. There are
not as many Asian and European readers there are in the US and
countries which share the US time zone which is an indication that a
fair fraction of the US east coast /. readership missed the initial
posting, but got it the following morning. (This is how one could
interpret the second rise in hits which peaks around 9:30am the
following morning.)
<p>
There is a second resurgence, on a much smaller scale, two day's later
which peaks at about 9:30am, with an abrupt fall off and then a
re-resurgence around 6pm that day. This is seen in Figure 4.
plot. The author scanned the various Linux news web pages for posting
of the /. effect article but found none. This small resurgence
occurring 2 days after the initial posting by /. is unexplained and open
to interpretation.
<p>
On or about midnight of the 3rd of Feb., Linux Weekly News posted an
announcement of the /. effect paper on one of its back pages. This is
indicated by the light blue time mark. There is an indication that
this posting contributed to the hit rate, at a much lower rate for
another 24 hours. This is best seen in the Figure 1, with the hit rate
following a gentle rise after the light blue time mark of the LWN
posting.
<p>
Finally, the small surge in hits before the /. posting, was due to the
author's announcing the availability of the paper to the redhat mailing
list as well as to some his colleagues via e-mail
<p>
<center>
<table>
<tr valign="top">
<td>
<table><tr><td><a href="./gx/adler/SDEPlot2.jpg"><img src="./gx/adler/SDEPlot2Small.jpg"></a>
<caption><font size=-1>
<center>Figure 2</center>
Data between 31/Jan/1999:20:00 and 02/Feb/1999:02:00. The small
surge in hit rate, around noon of that day, is due to the authors
announcing the paper to the redhat mailing list as well as to some
colleagues via e-mail.
</font></caption>
</table>
<td align="top">
<table><tr><td><a href="./gx/adler/SDEPlot3.jpg"><img src="./gx/adler/SDEPlot3Small.jpg"></a>
<caption><font size=-1>
<center>Figure 3</center>
Data between 01/Feb/1999:20:00 and 03/Feb/1999:02:00. This plot
contains the bulk of the hits recorded by the server. The initial it
rate, 15 minutes after the posting on /., is the highest. See body of
article for further details.
</font></caption>
</table>
<tr valign="top">
<td>
<table><tr><td><a href="./gx/adler/SDEPlot4.jpg"><img src="./gx/adler/SDEPlot4Small.jpg"></a>
<caption><font size=-1>
<center>Figure 4</center>
Data between 02/Feb/1999:20:00 and 04/Feb/1999:02:00.
The small resurgence in hit rate twice in this graph is unexplained.
</font></caption>
</table>
<td>
<table><tr><td><a href="./gx/adler/SDEPlot5.jpg"><img src="./gx/adler/SDEPlot5Small.jpg"></a>
<caption><font size=-1>
<center>Figure 5</center>
Data between 03/Feb/1999:20:00 and 05/Feb/1999:02:00. This plot
shows the low sustained hit rate occurring after the posting by
LWN.
</font></caption>
</table>
</table>
</center>
<p>
To conclude, this will be the last paper related to the Slashdot
effect. If for some reason this addendum to the /. effect generates
a new /. effect on the server, this will most likely not be presented
as an Addendum to an Addendum. Unless the /. wave front and ensuing hit
rate structure reveal new valuable data shedding light on the inner
workings of the Internet and the people who use it.
<!--===================================================================-->
<P> <hr> <P>
<center><H5>Copyright © 1999, Steve Adler <BR>
Published in Issue 38 of <i>Linux Gazette</i>, March 1999</H5></center>
<!--===================================================================-->
<P> <hr> <P>
<A HREF="./adler1.html"><IMG SRC="../gx/back2.gif"
ALT=" Back "></A>
<P> <hr> <P>
<!--startcut ==========================================================-->
</BODY>
</HTML>
<!--endcut ============================================================-->
|
