To evaluate user mobility, this study focus on the number of distinct APs visited daily by each user. Fig. 6.11 suggests that user mobility can be partitioned in 3 distinct periods.
6.2.3.1 2005 to 2007
The period between 2005 and 2007 is characterized by the stability of both the aver-age number of access points visited, and the session length. That is, the devices tend to
6.2. WIFI DATASET 77
be fixed, with no variations in the use pattern being observed during this period. This stability is consistent with the ratio of 1 to 1 between users and devices (cf. Fig. 6.2), the negligible weight of SMDs (Fig. 6.5) and the slow traffic growth rate suggested by Fig. 6.8.
It is also noticeable that the group of SMDs has a lower average number of visited APs, this is barely meaningful given that SMDs aren’t even 5% of the total number of devices. However the stability of this value suggests that the class of SMDs wasn’t as mobile as we could think of it, by this time most SMDs were PDAs (Personal Digital Assistants). This could suggest that the wireless technology at that time lacked the power efficiency of today wireless interfaces available on SMDs, having a large impact on battery, forcing users to turn on the wireless interface only under controlled use, turning it immediately off after using the application that required network access.
6.2.3.2 2007 to 2009
The second period (between 2007 and 2009) is characterised by an increase of nearly 100% of the average duration of sessions but without no change in the num-ber of visited access points per user. Fig. 6.8 shows that this period coincides with the increase of traffic, although the number of users continues to grow at an almost linear pace. The short distance between the number of devices and the number of users (cf.
Fig. 6.2), and the stability of the number of visited APs, suggests that this period is uniquely characterized by an increase in the volume of IPL eduroam network use.
6.2.3.3 2010 Onwards
The year of 2010 marks the beginning of a new use pattern where users connect to the network at a larger number of locations, although by shorter amounts of time.
In this period, the average session duration falls progressively to values that in 2013 are lower than the ones from 2005. Simultaneously, the average number of visited APs increases by more than 50%. This result confirms our expectation that a significant
0%
Figure 6.12: Distinct APs visited per user daily
Table 6.1: Maximum number of distinct APs visited by a user in a single day 2005 2006 2007 2008 2009 2010 2011 2012 2013
18 22 25 25 27 28 32 52 48
change is taking place in the wireless network usage pattern. This change is attributed to the wider deployment of SMDs, what is supported by noting that this is also the period in which the ratio between devices and users increases, and in which the ratio of SMDs gains relevance. It should also be noted that SMDs exhibit a pattern that dif-ferentiates from the average results, both when considering the number of visited APs (considerably more than the average) and the session duration (considerably shorter).
Figure 6.12 deepens the study on the mobility of users by analysing the distribu-tion of the number of access points visited daily per user. The results are in line with previous findings, with the values being stable until 2010. During this period, less that 10% of the users visited 4 or more APs in a single day. However, from 2010 onwards, it is possible to observe a steady growth in the number of users with more than 5 APs visited daily. In 2013, these were already more than 20% of the users. These results confirm the suspicions of an increasing user mobility. Although not statistically signif-icant, and mostly as curiosity, Tbl. 6.1 shows the evolution of the maximum number of distinct APs visited by a single user in a day.
On the other hand, the decrease in traffic and on the average session duration (resp.
6.2. WIFI DATASET 79
20%
30%
40%
50%
60%
70%
80%
90%
100%
< 1min < 5min < 15min < 30min < 1h <= 2h
2005 2006 2007 2008 2009 2010 2011 2012 2013
Figure 6.13: Detailed session duration
Figs. 6.8 and 6.11) raises questions about the usefulness of these sessions. Figure 6.13, which depicts the distribution of mean session duration, exhibits two distinct patterns.
Between 2005 and 2009, approximately 80% of the sessions had a duration of less than 5min, with more than 50% in the range between 1min and 5min. In contrast, from 2010 onwards, more than 50% of the sessions are shorter than 1min. Interestingly, the distribution of longer sessions is kept relatively stable across the study period.
The non negligible number of very short sessions (< 1min) can be interpreted in several ways. On one hand, it could result from protocol negotiation problems that force the devices to frequently restart the sessions. However, it would be expected that the proportion of sessions with problems remained constant over time, or even decreased as a result of technological improvements and error elimination in protocol implementations. Additionally, the attribution of these sessions exclusively to proto-col issues doesn’t justify the reduction in traffic, the increase on the number of daily visited access points by users or the decrease on the utilization time per user (Cf. resp.
Figs. 6.8, 6.12 and 6.10).
We claim that, instead, these very short sessions can be attributed to users that keep their SMDs with the radio interface enabled while moving between different campus locations, establishing connections with access points while on the move. This conclu-sion is supported by relating the two patterns (2005-2009 and 2010-2013) observed in
MobIPLity Bonnmotion Scenario RADIUS
Records
DHCP
Records ns-2
ns-3
OMNeT++
...
Figure 6.14: MobIPLity Work-flow
Fig. 6.13 with the introduction of SMDs, that since 2009 is gaining visibility. In addi-tion, Fig. 6.11 shows that the growth in the number of visited APs is not reflected in the session duration of SMDs. Interestingly, this could also hint changes in laptop use behaviour. Apparently, users are not replacing them by SMDs, but at the same time use laptops in new ways. These findings suggest a more judicious separation of activities between SMDs and laptops, with the former being used while on the move (thus the larger number of APs visited) and the later for other kinds of tasks for which SMDs are not suited.
In summary, these results confirm an increase in the number of users and devices, with the latter growing at a faster pace. Results also show that mobility has increased with users connecting to an increasing number of access points daily. The combination of both results suggests that there is an ongoing change in the use pattern of wireless networks. Users now tend to access to wireless networks through two devices, with one turned on even when the user is moving. However, the increase of the number of devices and connectivity is not reflected in traffic, which tends to decrease (in spite of an increase in mobile devices).