Google Cloud  (Source: Google)
Google dispatches a new feature called Cloud Dataflow Shuffle in its cloud service for customer data processing. This feature will enable the processing their data faster before they get to send it to a database or a machine learning system.

Google is putting together in making the analysis processing of data for cloud customer to become faster. With a new feature called Cloud Dataflow Shuffle, which is created to make batch processed data and consuming streaming as far as five times faster than it was previously, by the using a technology developed in-house.


This new feature is created for the Google Cloud Dataflow service that aids customers in processing data prior to sending it through to machine learning applications, databases and other systems. It works by enabling the customer in setting up processing tasks in the Cloud Dataflow directing it in pipelines that were written with the Apache Beam SDK, subsequently Google administers the materials and scaling of computer resources that will handle the tasks.


Photo source: Google


The Google Cloud Dataflow Shuffle dispatches the pipelines through the use of a Google made system to regulate shuffle operations. This sort data from various computer nodes, when activated customers will receive benefits at no additional cost. This is being made applicable as result of Google managing the Cloud Dataflow service, and will be able to switch in new components and features with the condition of it being possible and necessary. This new cloud feature, may benefit and appeal to customers that may choose to run Beam pipeline somewhere else in a different circumstance.


Photo source: Google


In addition to the SDK created by Google, it’s applicable for users to set up the pipelines on Apache Flink, Apex, Spark and Gearpump clusters executing in other locations. The significance of the Google Cloud Dataflow Shuffle depends on the magnitude at which a Beam pipeline depends on shuffle operations. Before one of Google data center came online, the company’s engineers once ran a 50,000TB shuffle on their servers for testing purposes.  As reported by William Vambenepe, a group product manager on the Google Cloud Platform team said that many of the pipelines that took the longest to run needed widespread use of shuffle operations.

Source: Flipboard

"Well, there may be a reason why they call them 'Mac' trucks! Windows machines will not be trucks." -- Microsoft CEO Steve Ballmer

Most Popular Articles

Copyright 2018 DailyTech LLC. - RSS Feed | Advertise | About Us | Ethics | FAQ | Terms, Conditions & Privacy Information | Kristopher Kubicki