The workload on parallel supercomputers: modeling the characteristics of rigid jobs

Lublin, Uri; Feitelson, Dror G.

doi:10.1016/s0743-7315(03)00108-4

Cited by 340 publications

(311 citation statements)

References 11 publications

Supporting

Mentioning

305

Contrasting

Unclassified

Order By: Relevance

“…As another example, Lublin has defined what may be called a "hyper log-gamma" distribution [453]. The log-gamma part comes from fitting log-transformed data to a gamma distribution.…”

Section: Do It Yourselfmentioning

confidence: 99%

“…The other approach is to use a parameterized distribution, and to model the daily cycle by showing how the parameters change with the time of day. In particular, interarrival times can be modified such that more arrivals occur during prime time, and less at night [320,453]. This first requires a model of the level of activity as a function of the time of day.…”

Section: Modeling the Daily/weekly Cyclementioning

confidence: 99%

“…It has a pronounced peak in the morning hours and a large dip at lunchtime. Lublin has proposed a simpler model, based on a gamma distribution, shifted so that the minimum occurs at 5 AM [453].…”

Section: Modeling the Daily/weekly Cyclementioning

confidence: 99%

“…One way to do so is to find a function that models the arrival rate as a function of the time of day [102,453]. This changes the requirement that the probability of an arrival event be the same for all intervals, and replaces it with different probabilities that reflect the different arrival rates.…”

Section: Nonhomogeneous Poisson Processmentioning

confidence: 99%

“…Such models are nontrivial because parallel jobs have several distinct attributes, most notably their parallelism and runtime, which have to be modeled. Several models and benchmarks for this domain have been devised [739,230,369,187,137,453]. They complement the benchmarks of multithreaded applications on modern multicore systems that were discussed in Section 9.2.2.…”

Section: Parallel Jobsmentioning

confidence: 99%

See 4 more Smart Citations

Workload Modeling for Computer Systems Performance Evaluation

Feitelson

2015

Self Cite

196

131

View full text Add to dashboard Cite

Reliable performance evaluations require the use of representative workloads. This is no easy task since modern computer systems and their workloads are complex, with many interrelated attributes and complicated structures. Experts often use sophisticated mathematics to analyze and describe workload models, making these models difficult for practitioners to grasp. This book aims to close this gap by emphasizing the intuition and the reasoning behind the definitions and derivations related to the workload models. It provides numerous examples from real production systems, with hundreds of graphs. Using this book, readers will be able to analyze collected workload data and clean it if necessary, derive statistical models that include skewed marginal distributions and correlations, and consider the need for generative models and feedback from the system. The descriptive statistics techniques covered are also useful for other domains.

show abstract

“…As another example, Lublin has defined what may be called a "hyper log-gamma" distribution [453]. The log-gamma part comes from fitting log-transformed data to a gamma distribution.…”

Section: Do It Yourselfmentioning

confidence: 99%

Section: Modeling the Daily/weekly Cyclementioning

confidence: 99%

Section: Modeling the Daily/weekly Cyclementioning

confidence: 99%

Section: Nonhomogeneous Poisson Processmentioning

confidence: 99%

Section: Parallel Jobsmentioning

confidence: 99%

See 3 more Smart Citations

Workload Modeling for Computer Systems Performance Evaluation

Feitelson

2015

Self Cite

196

131

View full text Add to dashboard Cite

show abstract

The Gridbus Middleware for Market‐Oriented Computing

Buyya

Venugopal

Ranjan

et al. 2009

Market‐Oriented Grid and Utility Computing

View full text Add to dashboard Cite

De‐centralized job scheduling on computational Grids using distributed backfilling

Wang

Gui

Zheng

et al. 2006

Concurrency and Computation

View full text Add to dashboard Cite

SUMMARYTo dynamically improve system selections of the waiting jobs under de-centralized scheduling frameworks, each computation system together with its neighbors is assumed to compose a subgrid, and distributed backfilling is proposed as the scheduling strategy in each subgrid. Whenever a job terminates, distributed backfilling is triggered to re-backfill all waiting jobs on the corresponding subgrid in order of their submittal. Each subgrid is overlapped with another, so the waiting jobs may be migrated around the Grid. To evaluate distributed backfilling, Grid resources and scheduling framework are configured, the model of Grid workload is constructed by extending the workload models of parallel systems and the Job Scheduling Simulation System is designed to simulate the process of de-centralized job scheduling. In addition, job speedup is presented as performance metrics of scheduling strategies. Results show the dynamic optimization of system selections brought by distributed backfilling is Grid-wide, and can improve scheduling performance remarkably as long as Grid load is not too light and the job migration costs are not too high.

show abstract

The workload on parallel supercomputers: modeling the characteristics of rigid jobs

Cited by 340 publications

References 11 publications

Workload Modeling for Computer Systems Performance Evaluation

Workload Modeling for Computer Systems Performance Evaluation

The Gridbus Middleware for Market‐Oriented Computing

De‐centralized job scheduling on computational Grids using distributed backfilling

Contact Info

Product

Resources

About