Review of Long tail

This article reveals that the long tail phenomenon, recently triggered by
on line services, is the fundamental nature of our world. It does not be
shown before due to limited storage resources and high storage cost. The
author predicts that long tail will become a new economic model, and he
encourages people to draw more attention on the unpopular items on the
request distribution curve since these objects could lead to a potential
large amount of profits beyond normal popular objects. This observation
sounds impressive, and it does reflect the demand of diversity tastes of
our human being.

Long tail is shown by on line service since they provide less limitation
about the resources. However, till now, the focus of computer system
designers is still on hotspots or popular items. It is well known that
Zipf's law is a popular request distribution in web requests. Although it
does reflect the long tail effect, the request percentage occupied by the
tail could be still small compared with popular items in many cases.

But this paper reminds us that the request distribution may obviate the
zipf's law if more and more attention of users will be put on the median
and unpopular objects. If such diversity is a trend, this observation will
change our design of replication strategies, and corresponding load
balancing strategies. Previous, the system designers often put most of
efforts to replicate hot objects, and relieve hotspots. Moreover,
mentioned in p2p workload paper, the popularity of objects in a file
sharing system may decrease quickly. This will also reduce the big impact
of popular items, and emphasize the importance of other unpopular items.

In brief, this is an interesting article to show us the existing and
potential impact of unpopular objects.
Received on Mon Nov 28 2005 - 11:09:31 EST