Blending randomness in closed queueing network models

Abstract

Random environments are stochastic models used to describe events occurring in the environment a system operates in. The goal is to describe events that affect performance and reliability such as breakdowns, repairs, or temporary degradations of resource capacities due to exogenous factors. Despite having been studied for decades, models that include both random environments and queueing networks remain difficult to analyse. To cope with this problem, we introduce the blending algorithm, a novel approximation for closed queueing network models in random environments. The algorithm seeks to obtain the stationary solution of the model by iteratively evaluating the dynamics of the system in between state changes of the environment. To make the approach scalable, the computation relies on a fluid approximation of the queueing network model. A validation study on 1,800 models shows that blending can save a significant amount of time compared to simulation, with an average accuracy that grows with the number of servers in each station. We also give an interpretation of this technique in terms of Laplace transforms and use this approach to determine convergence properties.

OptiSpot: minimizing application deployment cost using spot cloud resources

“…Moreover, the fluid-approximated models can be easily used with tools like LINE [18,23] to perform random environment analysis. Random environments are stochastic models used to describe events occurring in the environment a system operates in [6]. In our particular situation we model the random environment around spot price fluctuations, so to take into account their effect when computing the mean response time and the response time distribution.…”

“…Our new contribution is that we adopt fluid-approximated performance models [22], which can calculate response time distributions quickly enough to be used at run-time. We also use a random environment model [6] to represent the effects of external events to the system, which for now is limited to price fluctuations, but that can be easily extended to other events expressible as stochastic models. Finally, in our model we also consider the effects of having multiple CPUs in cloud resources (as it is the case for Amazon EC2) and the overhead due to load balancing in case of placement decisions that require resource replication.…”

“…The latter, which is referred to as random environment model [6], captures the stochastic nature of the operational environment, in which VMs can be lost and restarted as a result of spot price fluctuations and the consequent temporary switch to an on-demand pricing model. -the use of advanced fluid analysis techniques to accurately approximate response time percentiles, which are commonly used to constraint performance in servicelevel agreements, but which are usually hard to compute in queueing networks.…”

Towards a DevOps Approach for Software Quality Engineering

“…Here we focus on a QoS model based on Layered Queueing Networks [5,14], which can be evaluated with tools such as LQNS or LINE [11]. We consider the extended LQN model underlying LINE, which adds to the standard LQN a random environment to describe changes in the application beyond the control of the application manager [4]. A random environment can model for instance temporary VM failures or high-contention in virtualized deployments.…”

Line: Evaluating Software Applications in Unreliable Environments

“…This therefore limits the use of LQN models to tackle reliability-aware cloud resource provisioning, where, for instance, the virtualized environment can affect the application processing rates at very short time scales [11]. In addition, while existing methods and tools are effective in estimating mean performance metrics, the lack of support for the analytic computation of response time percentiles has been pointed out in the literature as a limitation of LQN models for SLO assessment [12].…”