Compare YARN vs. Standalone

- [Instructor] As we start working with Apache Spark,…it is interesting to see…how the resource coordinators differ…in their use of specific resources for Spark only clusters…as opposed to Spark and mixed use, say MapReduce.…So, for example, if you have a YARN cluster or a client,…your driver in a Standalone will run in the client,…but for the YARN cluster,…it will run on the application master.…If you have a YARN client, it'll run in the client.…

The resource requestor differs as well.…For Standalone, it'll run in the client.…For a YARN cluster or client,…it'll run in the application master,…and these are all JVM processes you might remember.…For the executor start, that will run in the Standalone…from a Spark Slave node.…In the YARN cluster or the YARN client,…it'll run from the YARN Node Manager JVM process.…Persistent services in the Standalone…will run from the Spark Master or Workers,…but in the YARN cluster or the client,…it'll run on YARN Resource or Node Managers.…

So, very big differences depending on…