Slow Performance of Amazon SQS compared with RabbitMQ

One thing to keep in mind here is that SQS is replicating your data across multiple AZ's. This is going to add to the time complexity compared to a single Rabbit or other MQ implementation.

If your single RabbitMQ instance goes down, are you ok with not being able to process messages or potentially losing data? If you are, you probably don't need replication or even disk persistence. But I'm guessing most use cases would care and thus, SQS offers a very hands off distributed MQ solution that is, in theory, insulated from a single point of failure.

According to this article you can run RabbitMQ on Amazon AWS (EC2) and it will be faster and possibly cheaper, there is a message per second and a financial comparison. There are some other howto guides on the RabbitMQ site