Page 37 - ITU Journal Future and evolving technologies – Volume 2 (2021), Issue 2

ITU Journal on Future and Evolving Technologies, Volume 2 (2021), Issue 2

At irst, we are reminded of the previously proposed In the model network, all users are moving according to
Distance Based Clustering Algorithm (DBCA) and Hier‑ the Random Direction Mobility Model (RDMM) [45]. In
archical Caching Algorithm (HCA) [34, 44]. DBCA aims RDMM users move towards a random direction with a
to form maximal cliques and boost cooperation among random speed until they reach the boundary of the area.
Helper Users (HUs). Moreover, HCA provides effective At the edge of the boundary users bounce back to the area
usage of users’ limited cache capacity and maximizes and keep moving until the end of the simulation. Since re‑
the network’s average of loading. Furthermore, we pro‑ alizing the cell borders are unlikely, bouncing users can
pose Advertisement Based (ABD2D) and Discovery Based be interpreted as newcomers to the cell. Fig. 1 illustrates
(DBD2D) D2D user/ content discovery procedures to the system model.
complement the DBCA and HCA. These schemes aim to
advertise cached content by the clusters and discovery of
cached content by Ordinary Users (OUs), respectively.
A model network comprises mobile cellular users where
some of them (HUs) are cache enabled and able to cre‑
ate clusters in order to make cache placement and con‑
tent discovery more ef icient, in terms of of loading and
control messaging overhead. Clustering and cache place‑
ment algorithms are applied consecutively before the pro‑
posed advertisement and discovery procedures. This is
the irst work that addresses the D2D‑based distributed
Fig. 1 – Illustration of Network Topology
cache placement and content delivery along with the crit‑
ical components of device discovery and content adver‑ Content set 1, 2, ..., is bounded and HUs have limited
tisement. We have implemented an LTE‑based simulation cache capacity ( ) which necessitates an excogitated

setup and analyzed the performance of proposed proce‑ cache placement approach. Each content is assumed to
dures in terms of average of loading and messaging over‑ be of equal size. Content popularity among users are uni‑
head against a varying number of Discovery Resource form, Zipf distributed at each user and assumed to be
Blocks (RBs), Helper Users(HUs), Ordinary Users (OUs) known by all users in the network. According to the Zipf’s
and Signal to Interference Noise Ratio (SINR) threshold. Law, out of a content set with elements popularity of
Results are presented, discussed and highlighted in Sec‑ the content of rank is de ined as [46];
tion 7.
1
( ∶ , ) = (1)
2. SYSTEM MODEL ∑ ( 1 )
=1
We consider a simplex content delivery procedure be‑
tween cellular users where HUs provide content to OUs
upon request. A model network consists of a Base Station Where istheskewnessexponentoftheZipfdistribution.
(BS) placedat thecenter ofacircularareaand all usersare Bigger reveals the domination of the most popular con‑
randomly distributed. Users are categorized into three tent, while on the other hand smaller induce to a more
groups: uniform‑likely distribution. is the size of the content
set and is the rank of the content where = 1 stands
• Helper Users (HUs) are able to cache content and for the most popular content of the set.
provide them to the OUs via D2D links upon request.
HUs are also able to form clusters with the help of For successful content delivery, in addition to content
the BS and manage intra‑cluster cache placement in availability, some other conditions must be met. First
a hierarchical structure. of all, users must meet the SINR threshold constraint to
exchange discovery/advertisement beacons. Since the
• Isolated Helper Users (IHUs) have the same role number of dedicated RBs for discovery and advertise‑
and abilities as HUs but at a certain time slot they ment procedures are limited and randomly selected by
have not been able to join any cluster and thus stay users, interference and collisions are inevitable. Without
alone. IHUs cache and serve content individually. exception, HUs and OUs, which are exposed to interfer‑
Since all HUs are mobile, relative positions are time‑ ence, are unable to establish D2D links in an Awareness
varying. Therefore, being isolated might be thought Zone (AZ), cannot receive and transmit content in the De‑
of as a transient state. livery Zone (DZ) and cannot make any contribution to of‑
loading sequentially. On the other hand, if discovery is
• Ordinary Users (OUs) are not cache enabled and ac‑ successful, enough resources are assumed to exist and
quire available content from HUs via D2D links. OUs content transmission is assumed to be successful. LTE
request desired content from BS if they cannot estab‑ physical layer and channel models [47] are adopted in the
lish D2D links via the proposed procedures. simulations.

© International Telecommunication Union, 2021 23