Information Technology Reference
In-Depth Information
some small losses rather than disastrous consequences. The workload PaaS platform
confronted fluctuates within a wide range. Over-provisioning system resources to
accommodate the potential peak will lead to the waste of the resource. As a conse-
quence, it is important for PaaS platform to maintain efficient resource utilization
under a wide workload conditions without increasing the possibility of failure of time-
liness. [9] proposed a new algorithm called FC-LRU which integrates feedback con-
trol with LRU algorithm to perform adaptive resource management in PaaS platform.
However, the idea of FC-LRU algorithm is based on the assumption that the PaaS
platform is using its dedicated infrastructure. Therefore, the resource management
mechanism in traditional PaaS platform is no longer practical in PoI. New mechanism
for PoI should be adopted.
3
Architecture of PaaS Based on IaaS
Some enterprises adopt PaaS to provide AEE for their applications and application
components, because PaaS is famous for optimizing development productivity and
business agility. However, the benefits of the traditional PaaS are limited, because the
cost of the infrastructure scale-down and scale-up is high and with low efficiency in
the typical PaaS. At the same time, IaaS has many advantages in optimizing infra-
structure utilization. In order to maximize business agility and development produc-
tivity, enterprises are trying to set up PaaS on IaaS, namely PoI. PoI provides applica-
tion development, testing, execution and provision environment which enables both
PaaS' ease of application deployment and IaaS' efficient infrastructure management.
In the traditional PaaS, the infrastructure is dedicated for its usage. In the idle
hours, some infrastructure is idle and lead to great waste; while in the busy hours, the
infrastructure is inadequate, and QoS of applications could not be guaranteed. In PoI,
when it is the idle hours, the infrastructure of PaaS could scale down by using the
capability of the underlying IaaS; when it is the busy hours, the infrastructure of PaaS
could scale up. This mechanism could greatly increase the Cost Efficiency of the
platform.
The architecture of PoI is shown in Figure 2. It can be divided into two layers: PaaS
Layer and IaaS Layer. The PaaS Layer dynamically rents VMs from the IaaS accord-
ing to the workload.
•
App Execution Engine
hosts users' applications. It provides multiple AEEs and
supports applications written by several programming languages (e.g., Java, PHP).
As one application is scaled up, the App Execute Engine will launch more in-
stances of this particular application.
•
Front End
is the entry to the platform: it accepts all requests from users and for-
wards them to the appropriate App Execution Engine. In essence,
Front End
is a
load balancer that knows the correspondence between requests and applications
.
•
Service Pool
manages a set of services including data storage and many other ca-
pabilities that can be used by developers via APIs.
