Bad news. Fail over clustering assumes a shared high perfomance storage device and Amazon doesn't provide that. The Amazon high perfomance storage device is the EBS volume, but you can attach it to only one EC2 instance. You will find Amazon S3 mentioned as a alternative all around, but it really isn't one. S3 is designed to persist data, and have very low IO perfomance.
I think the alternative is to setup a NFS, but a NFS is a higher layer protocol to share files without having a shared block device subsystem. It will work, but you will never get the absolutely best performance.
About EC2 vs RDS: EC2 will provide maximum performance if you want to tune every little aspect. RDS provides the best perfomance out of the box, everything is a click away, and you have no worries about maintaining the instance. Also RDS is easier to scale up and down as you need.
I found a MySQL benchmark, take a look.
Also, I found a SQL Server Whitepaper about implementing Microsoft Windows Server Failover Clustering (WSFC) and SQL Server 2012 AlwaysOn Availability Groups in the AWS Cloud: http://media.amazonwebservices.com/AWS_WSFC_SQL_Server_AlwaysOn.pdf