Databases Reference
In-Depth Information
A wide range and variety of applications can be deployed on a cloud infra-
structure. Consequently, data distribution and organization must be highly
e
cient. However, highly scalable data management schemes are in their in-
fancy. Our aim is to offer a solution for existing data management schemes
that are highly scalable and adaptive to dynamic changes in the applications
environment of the cloud.
Cloud systems overcome the limitations of existing grid implementations
by providing different levels of services in response to the users' requirements.
Cloud systems can be deployed on top of existing grid infrastructures by
providing an abstraction interface, which hides its complexities from the users.
It is our hope that the future of cloud computing includes greater utilization
of existing grid infrastructure.
6.1.3 Peer-to-Peer (P2P) Computing
The term “peer-to-peer (P2P) computing” was coined during the turn of the
twenty-first century and refers to a system that utilizes distributed resources to
perform functions in a fully decentralized manner. The P2P computing model
enables direct resource sharing between peers [77]. This capability makes a
system built on the P2P platform cost e
cient and effective. In grid comput-
ing, the participants might be clusters or high performance desktops that are
administered under a well-defined policy and trust. In contrast, a P2P com-
munity comprises diverse and anonymous participants with fewer restrictions.
A node can join and leave a P2P network without facing any bureaucracy,
and some P2P networks can scale up to hundreds of thousands or even mil-
lions of peers. More scalability means that more resources can be shared.
Unfortunately, this scalability comes with the trade-off of a dynamic network.
However, the peers in most of the available P2P systems are self-organized
and are able to recover automatically in the event of peer failure. A P2P
network operates without a server, i.e., all communicating peers are equal,
which reduces the risk of a single point of failure. The fault tolerance of the
system is ensured through data redundancy. A P2P network is more econom-
ical than a centralized or grid-based system because there is no server, which
is expensive to acquire and maintain, and the peers are not required to be
powerful machines with better connections, which must be maintained. How-
ever, P2P networks are challenged by data within the network because there
is no authority to monitor the network. Specifically, several issues arise in
file-sharing such as content pollution, out-of-control dissemination of sensitive
data, malicious peers, and copyright infringement.
P2P computing provides an alternative to existing client-server models. It
has a scalable resource sharing capability. Figure 6.3 compares a high-level
view of interconnectivity in a P2P computing model with a client-server ap-
proach. Current research efforts in P2P computing are focused on the devel-
opment of a dynamic file-sharing system. However, its real potential could lie
