vmBBProfiler: a black-box profiling approach to quantify sensitivity of virtual machines to shared cloud resources

Abstract

Virtualized Data Centers are packed with numerous web and cloud services nowadays. In such large infrastructures, providing reliable service platforms depends heavily on efficient sharing of physical machines (PMs) by virtual machines (VMs). To achieve efficient consolidation, performance degradation of co-located VMs must be correctly understood, modeled, and predicted. This work is a major step toward understanding such baffling phenomena by not only identifying, but also quantifying sensitivity of general purpose VMs to their demanded resources. vmBBProfiler, our proposed system in this work, is able to systematically profile behavior of any general purpose VM and calculate its sensitivity to system provided resources such as CPU, Memory, and Disk. vmBBProfiler is evaluated using 12 well-known benchmarks, varying from pure CPU/Mem/Disk VMs to mixtures of them, on three different PMs in our VMware-vSphere based private cloud. Extensive empirical results conducted over 1200 h of profiling prove the efficiency of our proposed models and solutions; it also opens doors for further research in this area.

vmBBThrPred: A Black-Box Throughput Predictor for Virtual Machines in Cloud Environments

“…That is, instead of challenging a VM to compete with other co-located VMs to access/use resources (CPU, Mem, and/or Disk), the vmLimiter limits resource usage of the VM so that it reveals its behavior under hypothetical contentions. We showed in [17] that although resource starvation under "contention" and "limitation" are different, they always lead to very similar performance degradation (less than 5% different on average). cpu/mem/diskLimit ∈ [0, 1] sets the percentage of CPU/Mem/Disk that the VM can use.…”

“…It is worth noting that each metric is a series of values during the profiling phase (eg. 15 values for 5 minutes of profiling in [17]), however because they showed to have negligible standard deviation, their average values proved to be accurate enough to be used in vmBBThrPred.…”

“…The key idea of vmBBThrPred is to use the sensitivity values of an application to model/formulate its throughout. vmBBProfiler, our systematic sensitivity analysis approach in [17], was designed to pressure an unknown VM to work under different working scenarios and reveal its sensitivity to each resource type. vmBBProfiler calculates three sensitivity values (∈ [0, 1]) upon profiling a VM: Sen c , Sen m , and Sen d to respectively reflect sensitivity of a VM to its CPU, Mem, and Disk.…”

“…We included this mode only to show that not only vmBBThrPred can be easily adopted/employed by current systems, but also its bundling with vmBBProfiler yields more than 95% accuracy in predicting throughout of any application with any resource sensitivity. Similar to vmBBProfiler [17], vmModeler also uses the normalized values of C, M, and D to propose a polynomial function with prototype…”

Heap Allocator: Uma Política de Alocação de Máquinas Virtuais em Ambiente de Computação em Nuvens baseada em Heap com Prioridade

“…Sua estratégia aloca máquinas virtuais que necessitam intensamente de um mesmo tipo de recurso em servidores diferentes. O autor recorreu ao conceito de sensibilidade formulado por [Taheri et al 2017]. A sensibilidade de uma máquina virtual por um recurso computacionalé diretamente proporcional ao nível que ela necessita desse recurso, ou seja, o quanto seu desempenhoé degradado caso esse recurso lhe seja disponibilizado de maneira insuficiente.…”