Model-driven computational sprinting

Morris, Nathaniel; Stewart, Christopher; Chen, Lydia Y.; Birke, Robert; Kelley, Jaimie

doi:10.1145/3190508.3190543

Cited by 30 publications

(15 citation statements)

References 39 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…We thus take a different approach. We observe that tasks tend to have fairly similar execution times [30], leading to an execution in waves. For instance, a job composed of 40 tasks executing in a cluster with 20 computing slots will start with a first wave of 20 tasks executing in parallel.…”

Section: Wave-level Modelmentioning

confidence: 80%

“…Each sprinting mechanism is controlled via a corresponding sprinting policy which determines what and when to sprint. Time-based policies levering timeouts to control the sprinting are rather common [23,30].…”

Section: Computational Sprintingmentioning

confidence: 99%

“…In addition, the job execution may include an initial setup time that lasts 1/µ k o time on average, and an intermediate shuffle Approximation ratio for reduce tasks in a priority-k job O Overhead stage M t Map stage with t map tasks left to process S Shuffle stage R u Reduce stage with u map tasks left to process stage that requires on average 1/µ k s time. We note that, when sprinting is enabled, the service rates can be approximately captured by the effective sprinting rates as a weighted average of the sprinted and non-sprinted execution times per task and class k. Predicting these rates is complex [30]. We assume that the effective sprinting rates are provided by an oracle for each class k and timeout value.…”

Section: Modeling Diasmentioning

confidence: 99%

“…Sprinting policies decide when and what to sprint, e.g., phases within job executions [43] and particular queries [25]. Several approaches have been explored in the single priority scenario, from simple heuristics [23], to queueing [32] and machine learning [30] models. DiAS extends the sprinting policies with multi-priority scheduling and approximation schemes such that the performance of all priorities can be improved.…”

Section: Related Workmentioning

confidence: 99%

See 3 more Smart Citations

Differential Approximation and Sprinting for Multi-Priority Big Data Engines

Birke

Rocha

Pérez

et al. 2019

Proceedings of the 20th International Middleware Conference

Self Cite

View full text Add to dashboard Cite

Today's big data clusters based on the MapReduce paradigm are capable of executing analysis jobs with multiple priorities, providing differential latency guarantees. Traces from production systems show that the latency advantage of highpriority jobs comes at the cost of severe latency degradation of low-priority jobs as well as daunting resource waste caused by repetitive eviction and re-execution of low-priority jobs. We advocate a new resource management design that exploits the idea of differential approximation and sprinting. The unique combination of approximation and sprinting avoids the eviction of low-priority jobs and its consequent latency degradation and resource waste. To this end, we designed, implemented and evaluated DiAS, an extension of the Spark processing engine to support deflate jobs by dropping tasks and to sprint jobs. Our experiments on scenarios with two and three priority classes indicate that DiAS achieves up to 90% and 60% latency reduction for low-and high-priority jobs, respectively. DiAS not only eliminates resource waste but also (surprisingly) lowers energy consumption up to 30% at only a marginal accuracy loss for low-priority jobs.

show abstract

Section: Wave-level Modelmentioning

confidence: 80%

Section: Computational Sprintingmentioning

confidence: 99%

Section: Modeling Diasmentioning

confidence: 99%

Section: Related Workmentioning

confidence: 99%

See 2 more Smart Citations

Differential Approximation and Sprinting for Multi-Priority Big Data Engines

Birke

Rocha

Pérez

et al. 2019

Proceedings of the 20th International Middleware Conference

Self Cite

View full text Add to dashboard Cite

show abstract

“…Also, the median percentage errors for all the cases are below 5.4%, 7.6% and 13.5% for PMD, Canneal and InceptionResnetV2, respectively. To showcase the total number of correct predictions made by the system, we leverage a CDF curve that has been used in the literature to showcase the effectiveness of the machine learning models [45]. Figure 10 illustrates that, for the PMD application, 80% of the predictions have error rates less than 9%.…”

Section: Validating the Accuracy Of The Performance Modelsmentioning

confidence: 99%

FECBench: A Holistic Interference-aware Approach for Application Performance Modeling

Barve

Shekhar

Chhokra

et al. 2019

2019 IEEE International Conference on Cloud Engineering (IC2E)

View full text Add to dashboard Cite

Services hosted in multi-tenant cloud platforms often encounter performance interference due to contention for non-partitionable resources, which in turn causes unpredictable behavior and degradation in application performance. To grapple with these problems and to define effective resource management solutions for their services, providers often must expend significant efforts and incur prohibitive costs in developing performance models of their services under a variety of interference scenarios on different hardware. This is a hard problem due to the wide range of possible co-located services and their workloads, and the growing heterogeneity in the runtime platforms including the use of fog and edge-based resources, not to mention the accidental complexity in performing application profiling under a variety of scenarios. To address these challenges, we present FECBench (Fog/Edge/Cloud Benchmarking), an open source framework comprising a set of 106 applications covering a wide range of application classes to guide providers in building performance interference prediction models for their services without incurring undue costs and efforts. Through the design of FECBench, we make the following contributions. First, we develop a technique to build resource stressors that can stress multiple system resources all at once in a controlled manner, which helps to gain insights into the impact of interference on an application's performance. Second, to overcome the need for exhaustive application profiling, FECBench intelligently uses the design of experiments (DoE) approach to enable users to build surrogate performance models of their services. Third, FECBench maintains an extensible knowledge base of application combinations that create resource stresses across the multidimensional resource design space. Empirical results using realworld scenarios to validate the efficacy of FECBench show that the predicted application performance has a median error of only 7.6% across all test cases, with 5.4% in the best case and 13.5% in the worst case.

show abstract