Analyzing Process Skew to Tune Queries

Shared-nothing MPP databases benefit from parallelized work when all the components can work equally. If there are specific components (node slices) that get far more work than others, the overall time to complete the job is impacted by the time it takes for the longest slice to finish it's work.  
Redshift provides underlying system tables that contain detailed information about the work done on each slices to execute the query.  
The Aginity Workbench for Redshift allows you to get quick access to this information and presents it in an easy way to visualize and analyze processing for skew and bottlenecks.  
To get to this feature, click the View menu and select Server Query History. Once you have used the table to locate the query of interest, you can right-click on the query and select Show Iterators for this Execution. You will then be presented with a view of the work occurring on each slice.

Have more questions? Submit a request


Please sign in to leave a comment.

** Aginity, Inc.’s Provision of Scripts and Similar Materials at Help Desk Center. For the convenience of Aginity Amp™ clients, we provide code snippets, scripts and similar materials at this Help Desk Center. Such materials are reference materials provided for illustration purposes only. These are intended to serve as an example for self-service clients and are generally geared to respond to common questions asked by similar clients. Such materials constitute Aginity’s intellectual property. Aginity Amp clients and their authorized users are permitted to use these materials in connection with their software license and/or subscription of Aginity Amp. Nothing herein shall limit Aginity’s right to use, develop, enhance, modify or market any of these materials as part of its business. These materials are not formally supported by Aginity or its affiliates. Usage of these materials does not guarantee any specific results, uptime, performance or error-free operation. Aginity disclaims all warranties of any kind, whether express, implied, statutory or otherwise, including any implied warranty of merchantability or fitness for a particular purpose.

Powered by Zendesk