I'd use it for running optical character recognition on millions of pages of government documents to make public information truly free, at http://citizenaudit.org/. But I really just need the 26 cores with a master node feeding documents for that, not actually map-reduce.

Hmm, if there's no MR involved, I'm not sure our GPUs would be very helpful for this type of job. What do you think?

I've recently became interested in Hadoop, map/reduce, HDFS. I would like to use it for learning purpose.

What types of jobs are you trying to implement?

Would like to use it primarily for configuration purpose. Admin/devops kinds of jobs.

Could you provide me your preferred email? Is it the one in your HN profile?

The one in my HN profile should be good. Thanks.

I have a Tera-butte-ton of security data that I could use hadoop for analysis. Have you a free cluster?

We've developed a parallel architecture using GPUs thereby reducing the cost by a few hundred factors. So, yes, we will have a freemium plan, and an affordable pricing model.

Email me at [email protected] if you're interested.

I might use it to analyze email conversations related to customer service.

What types of algorithms would you implement? Would they be compute bound?

How many cores, how much RAM, how much HDFS storage?

Our free tier will be:

26 CPU cores, 15 GB RAM, 60GB of SSD storage 1536 GPU cores, and 2GB RAM

for a certain amount of time. If you'd like a reserve instance, or more power/storage/throughput, we'll charge more accordingly.

Most typical usage of Hadoop I've seen (so far) is people dumping many terabytes of data into HDFS and then analyzing it. But maybe you're more focused on distributed computing than the "big data" side of it.