I just used it. I needed to do some analysis on Netflix streaming data (about 700 million rows in my dataset).

I had a bunch of text files that are basically logs, so I wrote a parer that injected or updated each row in DynamoDB and then put a queue item into an SQS queue to process that row if it was new (since the same ID could show up in any number of log files).

After the initial parsing I ran the other script to do the actual work by sucking items off the queue and processing the rows, which at that point had a column for each log line for that ID.