ERP: An elastic resource provisioning approach for cloud applications

Abstract

Elasticity is the key technique to provisioning resources dynamically in order to flexibly meet the users’ demand. Namely, the elasticity is aimed at meeting the demand at any time. However, the aforementioned approaches usually provision virtual machines (VMs) in a coarse-grained manner just by the CPU utilization. Actually, two or more elements are needed for the performance metric, including the CPU and the memory. It is challenging to determine a suitable threshold to efficiently scale the resources up or down. In this paper we present an elastic scaling framework that is implemented by the cloud layer model. First we propose the elastic resource provisioning (ERP) approach on the performance threshold. The proposed threshold is based on the Grey relational analysis (GRA) policy, including the CPU and the memory. Secondly, according to the fixed threshold, we scale up the resources from different granularities, such as in the physical machine level (PM-level) or virtual machine level (VM-level). In contrast, we scale down the resources and shut down the spare machines. Finally, we evaluate the effectiveness of the proposed approach in real workloads. The extensive experiments show that the ERP algorithm performs the elastic strategy efficiently by reducing the overhead and response time.

Characterization of task response time in fog enabled networks using queueing theory under different virtualization modes

“…The semiperfect virtualization mode is more flexible than the perfect virtualization mode. Here, the (15 fog cannot guarantee all arriving tasks the same service time, as was the case in the perfect virtualization mode. Rather, it guarantees service times that are uniformly distributed between pre-defined limits a and b, with a<b.…”

“…Let R R S , Q R S and S R S be the TR time, queueing time and service time, respectively, of remote tasks under the semiperfect virtualization mode. Given the fact that S R S is uniformly distributed, S R S ∼ U [a, b] , then the queueing model at the fog for this mode is an M/U/m system, with the first two moments of S R S being and Using ( 13), ( 14), ( 19), (20), and ( 12), we can find the expected TR time in the perfect virtualization mode to be where Using (15), the stability condition of the remote queueing system for the semiperfect virtualization mode is…”

“…Now that we have the distribution f S R B (t) at hand, the first moment of S R B is Using integration by parts, we get This result is logical, since without the speedup factor the expectation of a truncated exponential would be E[S R ] = τ + 1 µ . In a similar manner, utilizing integration by parts twice in the process, the second moment of S R B is Using ( 13), ( 14), ( 24), (25), and ( 12), we can find the expected TR time in the baseline virtualization mode to be where Using (15), the stability condition of the remote queueing system for the baseline virtualization mode is…”

“…Our study is novel in that it characterizes the TR time, using queueing theory, under three different virtualization modes: perfect virtualization, semiperfect virtualization and baseline. These modes can currently be implemented easily, as virtualization technology has advanced in recent years to the point that not only VMs can be placed [14] with the desired configuration, but also can be resized [15] up and down after placement to cope with changes in workloads. The resizing property, often called elasticity, is the crux of present work as each VM should be resized before receiving the next task based on the time requirements of that task.…”

“…CPU, memory, I/O and communications bandwidth. In [15] the authors present a framework for elastic VMs implemented by the cloud layer model, based on a Grey relational analysis (GRA) policy. The framework has the capability to provision resources required to yield a predefined computational power.…”