This project has moved. For the latest updates, please go here.

HydroDesktop + Azure Cloud? Storage requirements

May 19, 2010 at 8:55 PM

Hi all,

I'm working with Prof. Marty Humphrey here in the CS department of UVa, and we're looking at how one might run HydroDesktop in the Azure cloud.  Overall, it's an impressive project, kudos to the developers.  My question is if any of you, both developers and users, can give me some rough estimates of how much data a user might import/cache locally and work with in a typical usage scenario.  Are we talking 10GB?  1GB?  500GB?

I think one possibility of utilizing cloud (remote) resources is to simply expose a hydro-modelling web-service which the desktop program could access to upload data/query for modelling at large scale (maybe several hundred compute nodes if needed).  Thus, I'd like to know what kind of data volumes we'd be looking at.

Thanks in advance for any responses.  Again, a nice project overall, well done!

-Zach Hill

PhD Candidate
Department of Computer Science
University of Virginia
Charlottesville, VA

May 19, 2010 at 9:02 PM

Hi Zack,

Do you want to run the entire HydroDesktop in the cloud, or just expose modeling services to a desktop version of HydroDesktop?  I think the 2nd option is a better route, personally.  The amount of data that would need to be passed between client and service would, in most cases, be <1GB (of course there are always exceptions).

By the way, we met a few weeks back when I visited UVA.  Hope all is well.

Jon Goodall
U. of South Carolina

May 19, 2010 at 9:24 PM

Hi Jon,

Thanks for the response.  I agree that the 2nd option, a web-service for modeling, seems the best option particularly given how much work the HydroDesktop team has put into the nice GUI and local desktop tools.  <1GB sounds reasonable.  I've been looking at the code and the various plug-ins and it seems the architecture would lend itself to a hybrid local+cloud solution well.  The Hydro-Forecaster is an interesting component.  I'm not exactly sure how it's used but it seems like the logical component to move into the cloud (the actual computation of the model and workflow generated specifically).  Is that the primary 'modeling' component or have I missed something?

 

Yes, I remember you well.  Good to hear from you.

 

-Zach

May 20, 2010 at 1:50 PM

Zack,

I am not familiar with Hydro-Forecaster.  I am developing HydroModeler and, to the best of my knowledge, it is the primary modeling plug-in for HydroDesktop (others, please correct me if I am wrong).  That said, the HydroDesktop architecture is very open, as you noted, and so anyone can develop additional plug-ins to suit their needs.

Jon

 

Coordinator
May 20, 2010 at 3:13 PM

HydroForecaster was an experimental tool one of my students was working on using our Model Builder like workflow environment. Jon and I are hoping to figure out if it's possible to merge the OpenMI workflow GUI with our Model Builder one. But that will take some time...