Performance Analysis of Cloud Computing Centers Using M/G/m/m+r Queuing Systems

A Specification Language for Performance and Economical Analysis of Short Term Data Intensive Energy Management Services

“…The distribution of response time for a cloud modeled as a queue system when service time is not exponential is complex and rely in some approximation. These approximations are very sensitive to the probability distribution of task service times, and they become increasingly inaccurate when the Coefficient of Variation (CoV) increase towards the value of 1 [17,7,10]. The difficulty increases when we try to analyze performance of applications implementing advanced data flow abstraction with a high level of concurrency constraints.…”

On the Performance of the Spotify Backend

“…Extensive research has been undertaken in modeling the performance of storage devices [16][17][18][19][20][21]. The authors in [16] present a performance model for Ursa Minor [22], a robust distributed storage system.…”

“…In [20], a queuing model is developed to model response times of a cluster of servers, considering general service times. The model gives the response time in a transformed form, or by its moments, and further approximations would be needed to derive its distribution.…”

“…In [21] the service time distribution in clustered video servers is evaluated, approximating non-dominant delay components with constant values. Note that the models in [19] and [20] are evaluated on a simulator only, and the works do not include testbed experiments.…”

Untitled

“…These service times include computation as well as access latencies associated with database access and network communications. Such assumptions are commonly used in analyzing cloud computing systems (Ali-Eldin et al, 2012a;Wang et al, 2008;Urgaonkar et al, 2008;Urgaonkar, Pacifici, Shenoy, Spreitzer, & Tantawi, 2005;Khazaei et al, 2012;Kikuchi & Matsumoto, 2011;Firdhous, Ghazali, & Hassan, 2011). We also assume a Poisson arrival for incoming jobs.…”