6

I sometimes have some data processing jobs to run, but my laptop is not fast enough (2.4 Ghz Core 2 Duo) or don't have enough memory (4 GB). It may need to run for days. I want to buy fast computing time on demand. Here is my ideal workflow:

1) Write my code in python, R or C++.

2) Connect to the server which has persistent storage of the libraries I installed before

3) Upload my code and data to the server

4) Run it much faster than my laptop

5) Download the output file

6) Disconnect and pay very little

QUESTION: Where is the best (cheap and fast) place do this?

3
  • I think this would depend heavily on what you define as 'very little' in point 6. Presumably you're looking at some form of Cloud Computing - in which case rates tend to start in the region of $0.10 per hour or so (Azure / AWS for example).
    – James Gaunt
    Nov 10, 2010 at 0:35
  • This falls into one of those tricky gray areas in terms of on-/off-topicness. I've voted for migration to Server Fault simply because I think you'll find more experience there. Nov 10, 2010 at 0:52
  • In my experience, RS cloud instances have a lot more CPU per $ than EC2. Oct 16, 2013 at 2:44

6 Answers 6

4

I don't know if it fits your needs but I really dig the http://www.picloud.com/ idea.

3

Amazon EC2 has some HPC (high performance computing) offerings

1

I tried Amazon AWS and Rackspace. Both are great.

0

The cloud service that gives you the most CPU for your dollar is probably Linode; Amazon is notoriously poor on this benchmark. The drawback to Linode, however, is that you pay by the month, while with Amazon you can pay by the hour. I'm not familiar with Rackspace's offering.

0

Joyent seem to brag they're better than Amazon EC2 in performance and cost, but I've no personal experience with their setup yet:

http://www.joyent.com/services/cloudhosting/joyent-vs-amazon-ec2/

0

I use http://prgmr.com. You pay monthly, but there are no contracts so you could just decide to pay a month or two and be done.