Entering edit mode
6.3 years ago
Shicheng Guo
★
9.6k
TACC vs SDSC vs HTCondor ....
Any senior research can give a better comparison between them. Or Is there any other good computational service to choose for big data analysis?
I used SDSC before. Buying the machines on SDSC isn't cheap, it's like $10k per node, but they do offer a good ideal on storage called universal storage which costs roughly $1000 per month for a minimum of 200TB. But yeah, for hosting private nodes in SDSC you have to beware of the extra cost of maintenance for hiring SDSC consultants. For shared compute infrastructure in SDSC, you have to deal with the fact where you don't have root permission, which can be a pain when you need it urgently.
Anyhow, I highly recommend using resumable spot instances from AWS and setting the max bid price to 1/5 of the on-demand-cost. It's probably the least problematic solution that yields the most availability with lowest cost. Most labs I know are migrating to the AWS now.
I have written a blog for this kind of things, feel free to check it if you are interested: https://brianyiktaktsui.wordpress.com/2018/08/11/buying-computing-infrastructure-vs-adopting-the-cloud/
Also, buying in and opting out with AWS is way simpler. You can do it online. All you need is a credit card to try out AWS. The computing centers you mentioned sounds like the ones where you have to communicate back and forth through email.
I used all of them and SDSC actually has the best usage experience