Ensure that disk space and memory are not exceeded, and compute is not pegged at 100. Ensure that the Redshift cluster is sized appropriately.Note: This section refers to sort keys on the underlying table in Redshift, not the Looker parameter of the same name. Leverage interleaved sort keys for complex data sets (such as sessionized data). Utilize date fields, join fields, and/or fields that are commonly used as filters for reporting as sort keys. Set strategic sort keys on underlying database tables.If the skew is high, then common joins and expressions are likely under‐utilizing some nodes, while others are overworked.
Monitor the skew of the cluster when using distribution keys.Only consider "even" distribution style if a table is really large and never (or very rarely) needs to be joined to another table, as this will ensure that the power of the entire cluster is used for computing, and the need for colocated data is minimal when no joins are occurring.This will cause the table to be present on all nodes and therefore always colocated with other data required in a query. Consider utilizing the "all" distribution style for very small tables that are frequently used in queries.Note: This refers to distribution keys on the underlying table in Redshift, not the Looker parameter of the same name. Using the same distribution key for multiple tables causes data to be colocated across nodes, reducing network traffic and improving query performance. A distribution key dictates how data gets distributed across the nodes of a cluster. Set strategic distribution keys on underlying database tables.Denormalize Redshift schemas where possible, to reduce the number of joins required, improving parallelization and therefore query performance.There are a variety of things that you can do to reduce network traffic and improve query performance: While some network traffic is inevitable, reducing network traffic can help mitigate this performance degradation. However, query performance is negatively affected when too much data has to be transferred across a cluster network. The practices are written to work for most users and situations, but as always use best judgment when implementing.Īmazon Redshift is a clustered, columnar-store cloud database, that consists of nodes and is well‐suited to large analytical queries against massive datasets. These insights come from years of experience working with Looker customers from implementation to long-term success. These best practices reflect recommendations shared by a cross-functional team of seasoned Lookers. Looker will not be updating this content, nor guarantees that everything is up-to-date.
0 kommentar(er)
