Yes! You can find data that should be useful to you on clouddata.netapp.com . The graphs on clouddata are produced by running a simplified artificial workload on various types of storage including Amazon EBS and Ontap Cloud - and are updated every four hours. In the upper right corner there is a Sign-In button where you can sign in with SSO credentials. If you don't sign in, some of the graphs will be old, but if you sign in, then all the graphs will be the most recent ones. The information provided here is slightly different from what you want because the workloads are run via iSCSI to block devices. You were asking about NFS on EBS, note that this is different from Amazon's new EFS offering which is on a different stack. So you can use the information on clouddata to say "This is the latency of the raw EBS gp2 stack. Adding NFS on top of the EBS may change this number slightly, but it should be close to this." And the same can be said for Cloud Ontap. "This is the latency of raw disk from Cloud Ontap. Using NAS instead of SAN may change this number slightly, but it should be close to this." You want to compare the graphs labeled: "AWS EBS GP2 Maximum Latency" and "AWS COT Maximum Latency". Note that these are 100% random read workloads. Not a very realistic workload, but it shows the far end of the read spectrum. We do also have other workloads on that page for the COT. We collect those workloads for AWS EBS GP2, but they are not shown on the main page. We also collect this data for NPS both as spinning disk FAS and as AFF.
Yes, this occurs because the graphs you are looking at show a 100% read workload, and ONTAP Cloud provides caching, which means that some reads come back much faster. EBS GP2 has no caching, so they do not get this benefit. I've created some slides that show this phenomenon better, which I could email to you if you like.