Splunk Analytics for Hadoop will work with either your own hosted Hadoop cluster in GCP or Google Cloud Dataproc. You can access the cluster from a search head in GCP or external to GCP. The key is to make sure all the required ports are open for communication; 8020 for HDFS, 8032 for the Yarn resource manager and 8030 for the Yarn scheduler. The ports may vary depending on your setup or distribution, but those are the default ports for Dataproc. The same system requirements apply for either deployment.
I'll be publishing a how-to Blog post on access Dataproc with Splunk Analytics for Hadoop soon. I'll update this thread when it's available.