CherryPick: Adaptively Unearthing the Best Cloud Configurations for Big Data Analytics

By Omid Alipourfard, Hongqiang Harry Liu, Jianshu Chen, Shivaram Venkataraman, Minlan Yu, Ming Zhang. Symposium on Networked Systems Design and Implementation, 2017

523 citations recorded by the source corpus.

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

Autopilot: workload autoscaling at Google (2020, background): “h is equal to h[1] = 1 · 9 (load of 1 during 9 time units) and h[10] = 10 · 10 (load of 10 during 1 time unit).”
A generic communication scheduler for distributed DNN training acceleration (2019, methodology): “At each given (δ , c), the objective function value follows a distribution and we use Gaussian as it is widely accepted as a good surrogate model for BO [8].”
ProBO: Versatile Bayesian Optimization Using Any Probabilistic Programming Language (2019, methodology): “We can observe this pattern for a query, and use it to assign a state [1, 8].”
Pocket: Elastic Ephemeral Storage for Serverless Analytics (2018, background): “CloudScale [68], Elastisizer [40], CherryPick [11], and other systems [73, 77, 48] take an application-centric view to rightsize a job at the coarse granularity of traditional VMs as opposed to determining fine-grain storage requirements.”
Chameleon: scalable adaptation of video analytics (2018, methodology): “Ernest [31] uses optimal experimental design [28] to optimize the VM configuration of a job, while Cherrypick [12] uses Bayesian optimization [27] to find an optimal cloud configuration for general applications.”
Optimus: an efficient dynamic resource scheduler for deep learning clusters (2018, background): “, FABOLAS [47], BOAT [28], CherryPick [24]) to search best hyperparameters/resource configuration for a model/job.”