Information Technology Reference
In-Depth Information
Number of bytes transferred, or byte count.
Overall hit count defined above
can be misleading if the workload represented by individual hits shows high
variability. Consequently, we can choose the number of bytes transferred, or
byte count, as the workload measure of finer granularity, which can be easily
obtained by counting the number of bytes transferred for each hit recorded in
access logs.
Number of users, or user count.
User count is another alternative workload mea-
sure meaningful to the organizations that maintain the web sites and support
various services at the user level. When calculating the number of users for
each day, we treat each unique IP address as one user. So, no matter how many
hits were made from the same computer, they are considered to be made by
the same user. This measure gives us a rough picture of the overall workload
handled by the web site.
Number of user sessions, or session count.
One of the drawbacks of user count
is its coarse granularity, which can be refined by counting the number of user
sessions. In this case, along with the IP address, access time can be used to
calculate user sessions: If there is a significant gap between successive hits from
the same IP address, we count the later one as a new session. In practice, the
gap size can be adjusted to better reflect appropriate session identification for
the specific types of web applications.
The number of user sessions per day may be a better measure of overall work-
load than the number of users, because big access gaps are typically associated with
changes of users or non-web related activities by the same user. Each user who ac-
cesses the same web site from the same computer over successive intervals will be
counted by user sessions, as long as such a gap exists in between. Even for a single
user, a significant access gap is more likely to be associated with different usage pat-
terns than within a single time burst. Therefore, by using user sessions, we can count
the users' active contribution to the overall web site workload more accurately.
To summarize, the above measures give us workload characterization at different
levels of granularity and from different perspectives. Hit count is more meaningful
to web users as they see individual pages; byte count measures web traffic better;
while number of users or sessions provide high-level workload information to web
site hosts and content providers.
6 . 2 M e a s u r e m e n t R e s u l t s a n d W o r k l o a d C h a r a c t e r i z a t i o n
On the average, each day for the SMU/SEAS web site is associated with
301.6 Mbytes, 29,345 hits, 2338 sessions, and 2120 users; each user is associated
with 13.8 hits; each user session is associated with 11.6 hits; and each hit is asso-
