Information Technology Reference
In-Depth Information
constraints on the response time, but the failure of timeliness will lead to some small
losses rather than disastrous consequences. In addition, computing resources that
PaaS needs are provided by IaaS Layer as a set of VMs. For those enterprises willing
to adopt PoI, it is important to minimize the usage of VMs in order to minimize the
rental cost as well as guarantee QoS of their applications.
QoS of the application is measured based on the number of requests missed their
deadlines. Missed Deadline Ratio (denoted as MissRatio) is defined as the percentage
of requests that missed their deadlines; the formula used to calculate this value is
given by (1):
MissRatio
number of requests missing their deadlines
number of total requests
1
In terms of resource utilization, CPU utilization is involved in this paper because
CPU is the scarcest computing resource. It is important for PoI to maintain CPU utili-
zation at a high level. Different instances of an application may have different CPU
utilization, the CPU utilization of the application (denoted as
U
) is defined as the
average of CPU utilization of all instances; the formula used to calculate this value is
given by (2):
nM
∑ U
2
U
CPU
Where n is the number of instances. M is the maximum CPU utilization that each
instance is allowed to use (CPU utilization of one application's instance can be limited
by cgroup in recently released Linux version). M means that CPU utilization of an
instance should be limited below the M (e.g., 20%).
U
CPU
is used to represent the
CPU utilization of
i
instance of application j. The overall CPU utilization of PaaS can
be relatively high if the CPU utilization of each application is high.
Since users' applications are running in VMs provided by IaaS, it is equally impor-
tant to reduce the usage of VMs in order to lower the rental cost as much as possible.
Traditionally IaaS providers charge according to Pay-As-You-Go pricing model.
Therefore the rental cost is determined by how long and how many VMs are used. The
Usage of VM is defined as the total usage time (denoted as
T
VM
); the formula used to
calculate this value is given by (3):
3
T
VM
T
VM
Where
T
VM
represents the usage time of
i
VM and n represents the number of
rented VMs.
In PoI, the resource management framework is designed following three objectives:
•
Guaranteed QoS of the application, described by
Missed Deadline Ratio
.
•
High resource utilization rate, described by
CPU Utilization
.
•
Less VM usage, described by
Time Usage
of VMs.
