Performance modelling of parallel computer architectures

Harrison, Peter G.; Field, A. J.

doi:10.1145/317531.317535

Cited by 45 publications

(76 citation statements)

References 6 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…Similarly, this paper uses io q to denote the probability of the task that finishes it services at web server i and leaves the web cluster. From [5], [8], [9], [10], [11], and [12], the traffic equation for the network is given by:…”

Section: This Paper Usesmentioning

confidence: 99%

“…However, this model, the web servers do not communicate to each other. Therefore, [7], [10], and [14] that the average waiting time in the network is given by:…”

Section: This Paper Usesmentioning

confidence: 99%

“…Table I summarizes the system and workload parameters that compare the performance of the minimum entropy load balancing strategy with the traditional RR-DNS, the RR-DNS [3] and [10]. HTML document response time is a major index to measure the performance of the load balancing strategy.…”

Section: Parameter and Experiments Designmentioning

confidence: 99%

See 2 more Smart Citations

An Analytical Model of Web Server Load Distribution by Applying a Minimum Entropy Strategy

Nandhakwang¹,

Malisuwan²,

Sivaraks³

et al. 2013

IJCCE

View full text Add to dashboard Cite

Abstract-This paper presents an analytical model and the way of simulation for distributing workload on a distributed web server system. The increase in the Internet traffic has also necessitated the conventional Domain Naming Service (DNS) to operate at a much lower efficiency. Among a number of problems associated with the DNS, a key problem has to do with the authoritative DNS not being able to process complete knowledge of the proximity. This makes the authoritative DNS less effective in monitoring server availability and routing incoming requests around failed servers. The workload distribution strategy on the other hand, keeps track of the state and health of the web server. This avoids connection delay due to, for example, a failed server, which can be temporarily by passed by workload distribution. From a modeling standpoint, the conventional DNS assumes equal queue size for each web server in a round-robin setting. Under load balancing, the queue size for each web server differs based on the probability of accessing that server. This probability is based on such factors as the geography, server health, server response, server threshold, session capacities, and the round trip time. In this paper, both conventional and global workload distribution strategies are developed and compared based on a finite set of practical traffic scenarios.Index Terms-Domain name server, entropy, workload distribution strategy, round-robin setting.

show abstract

Section: This Paper Usesmentioning

confidence: 99%

“…However, this model, the web servers do not communicate to each other. Therefore, [7], [10], and [14] that the average waiting time in the network is given by:…”

Section: This Paper Usesmentioning

confidence: 99%

See 1 more Smart Citation

An Analytical Model of Web Server Load Distribution by Applying a Minimum Entropy Strategy

Nandhakwang¹,

Malisuwan²,

Sivaraks³

et al. 2013

IJCCE

View full text Add to dashboard Cite

show abstract

“…">IntroductionPerformance evaluation of information-system development becomes more complex as the size and complexity of the system increases (c.f., [7,11,17]). Reliability is certainly one of the most important characteristics for communication networks.…”

mentioning

confidence: 99%

Reliability analysis of complex communication systems

Sztrik

2000

J Math Sci

View full text Add to dashboard Cite

IntroductionPerformance evaluation of information-system development becomes more complex as the size and complexity of the system increases (c.f., [7,11,17]). Reliability is certainly one of the most important characteristics for communication networks. The measure of greatest interest is the distribution of the time to the first system failure. It is well known that the majority of problems can be treated with the help of serni-Markov processes (SMP). Since the failure-free operation time of the system corresponds to sojourn time problems, we can use the results obtained for SMP. If the exit from a given subset of the state space is a "rare" event, that is, it occurs with a small probability, it is natural to investigate the asymptotic behavior of the sojourn time in that subspace (see [5,6,8,10]). Realistic consideration of certain stochastic systems, however, often requires the introduction of a random environment, sometimes referred as to Markov-modulation, where system parameters are subjected to randomly occurring fluctuations. This situation may be attributed to certain changes in the physical environment such as weather, or sudden personal changes and work load alterations. In [4], Gaver proposee an efficient computational approach for the analysis of a generalized structure involving finite-state-space birth-and-death processes in a Markovian environment. Stern and Elwalid in [12] used Markov-modulated processes for analyzing some infinite-source information systems. This paper deals with a first-come, first-served (FCFS) queueing model to analyze the asymptotic behavior of a heterogeneous finite-source communication system with a single processor. The sources and the processor are assumed to operate in independent random environments. Each message is characterized by its own exponentially distributed source and processing time with parameter, depending on the state of the corresponding environment. Assuming that the arrival rates of the messages are many times greater than their service rates ("fast" arrival), it is shown that the time to the first system failure converges in distribution, under appropriate norming, to an exponentially distributed random variable. Some simple examples are considered to illustrate the effectiveness of the method proposed by comparing the approximate characteristics to the exact ones. This paper generalizes the results of Sztrik and Lukashuk [15] and Sztrik and Rigo [16], where the sources are homogeneous and the whole system is governed by a single random environment and two random environments, respectively. The technique used here is similar to the one applied in [9] and [13,14]. Preliminary ResultsIn this section, a brief survey is given of the most related theoretical results, mainly due to Anisimov (see ]1-3]), to be applied later on.9 Let (Xc(k), k _> 0) be a Markov chain with state space rn+l Uxq, x, nxj=o, i#j, q=O defined by the transition matrix (pc(i (q), j(':))), satisfying the following conditions:

show abstract

“…max t 2 min β The Laplace transform of the disk's response time pdf is derived using the Pollaczek-Khintchine transform equation for M/G/1 queues [10]:…”

Section: Data Transfer Timementioning

confidence: 99%

A Response Time Distribution Model for Zoned RAID

Lebrecht

Dingle

Knottenbelt

Analytical and Stochastic Modeling Techniques and Applications

View full text Add to dashboard Cite

Abstract. RAID systems are widely deployed, both as standalone storage solutions and as the building blocks of modern virtualised storage platforms. An accurate model of RAID system performance is therefore critical to understanding storage system performance. To this end, this paper presents a queueing network-based model of RAID systems comprised of zoned disks and operating at RAID level 0-1 or 5. The contribution over previous work is twofold. Firstly, our analysis approximates full I/O request response time distributions rather than just mean values. This provides the ability to reason about response time quantiles and higher moments of response time -both of which are useful in the context of modern quality of service requirements. Secondly, we validate our model against measurements from a real RAID system rather than a software simulation. The close agreement between predicted and observed response time distributions gives a high level of confidence in the validity of our model.

show abstract

Performance modelling of parallel computer architectures

Abstract: In this paper we describe two types of complex server aggregations which can be used to model collections of components in certain types of parallel computer systems and give a case study showing how the aggregations may be applied in practice.

Cited by 45 publications

References 6 publications

An Analytical Model of Web Server Load Distribution by Applying a Minimum Entropy Strategy

An Analytical Model of Web Server Load Distribution by Applying a Minimum Entropy Strategy

Reliability analysis of complex communication systems

A Response Time Distribution Model for Zoned RAID

Contact Info

Product

Resources

About