Yes, it works great for science calculations. I used about 15 machines or so to make a big quartet tree a while ago and it worked great. [1] The two main problems I remember with Amazon EC2 were the potentially changing dynamic IP address assignments and the chance that you can lose all data if you terminate your instance. But I think both of these are mitigated now. Such as with S3 as a storage backend. Cheers,
Rudi [1]: http://www.nature.com/nature/journal/v444/n7119/full/444528a.html On Tue, Feb 26, 2008 at 6:51 AM, Steffen Moeller <[EMAIL PROTECTED]> wrote: > Hello, > > you may have heard of the compute cluster of Amazon that everyone can > use on a per-cpu-minute-basis. It is pretty amazing. 1000 CPU hours is > 100$ or in other words if you exchange your 100% busy 500$ zero-energy > machine every 7 months then you are better of with Amazon. > > I found a Debian Etch image > http://developer.amazonwebservices.com/connect/entry.jspa?externalID=639 > > and some Ubuntu with minimal Bioinformatics > > http://developer.amazonwebservices.com/connect/entry.jspa?externalID=1259&categoryID=101 > > I cannot tell exactly about the difficulties that arise with an image > that comprises the complete Debian-Med. However, at least for docking or > molecular dynamic with gromacs, I could well imagine that this would > make perfect sense to try. > > I will not need it myself in the next couple of months, but if someone > of yours is thinking about hiring an extra technician to help a cluster, > then please be aware that 50k$ would pay for 500kh which is 57 CPU-years > alone. What I personally like is that one can easily specify costs for > cluster-computations in grant proposals. > > Best, > > Steffen > > > -- > To UNSUBSCRIBE, email to [EMAIL PROTECTED] > with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED] > > -- Which is worse, ignorance or apathy? Who knows? Who cares? --Erich Schubert -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
