Hacker News Reader

Comment by milancurcic | Hacker News Reader

Cofounder of Cloudrun here.

Cloudrun is a SaaS platform for the Weather Research and Forecasting (WRF) model. WRF is by far the most widely used numerical weather prediction model, but it currently requires access to supercomputers and domain expertise. Our goal is to make custom weather prediction possible for those without access to these resources.

This is our first MVP and we're working on making it more accessible for non-modelers. We greatly appreciate any feedback on how to make this service easier to use!

View on HN

This is _literally_ a cloud service :)

Bad joke aside: I'm curious, what is the "business" need and market/end-users?

Congrats and good luck Milan and Josh.

Thank you! We encourage bad jokes at Cloudrun. Our next project thedarkweb.edu is a site for custom-made blackout curtains.

Our end-users are currently mostly students and those working in academia. For these gals and guys, it can be a PITA to use supercomputers to run weather models. Milan can speak more to this.

However, the potential user is anyone who needs custom weather prediction. Renewable energy, commodities trading, event planning (sports, concerts, etc.) to name a few.

We're hearing about new potential industries all the time! Ideas are welcome!

Not sure if you envision getting down to a consumer level in some way, but I race sailboats. There's plenty of people who use forecasting services to get a forecast tailored to a very small area for their race course or route (think 10 sq miles).

I'd presume that mountain bikers, kayakers, climbers, and really anyone else who is into sports that are dependent on weather may be interested in similar information.

Delivering the information to a community like that seems like it would be too niche for your service, but for the people that already offer services to those people, perhaps there'd be interest.

Thank you for the feedback! At this point we don't target end consumers, but rather small businesses that create these kind of products. However, we do get a lot of requests for specialized and intuitive graphics/visualization and may go further in that direction later this year.

Naive question from a semi-beginner about your market: isn't there already a lot of competition for hyperlocal forecast and good accessible visualizations (e.g., darksky)?

Most existing services that I am aware of provide nice visualization of existing publicly available datasets (GFS, NAM, HRRR, etc.). Cloudrun allows the user to create their own forecast product/dataset, so it is mainly a power tool and less a graphics/visualization service.

Thank you, mino!

At this point, we are targeting existing WRF users and newcomers into the field. These are either academic meteorologists without access to big hardware, or SMBs with focus on weather intelligence for agriculture, commodities and energy trading, and similar.

At a later stage, we will probably move toward operational weather prediction, and expand to ocean circulation (ship routing, search & rescue) and surface wave models (custom surf forecasts).

As Josh said, though, we've just started and learning every day about potential use cases.

Do you supply all cores (NMM-B & ARW)?

How about all the other configurations of WRF (HWRF, WRF-DA, WRF-CHEM, Global)?

What sizes can you scale to? What type of hardware are you running on, if you don't mind me asking?

Thanks.

For now, we only offer ARW in real data case mode (em_real). We will slowly enable other modes and WRF components if there is demand.

All runs go on Intel Xeon Platinum 8168. We currently restrict the execution to shared-memory mode while we fine-tune the scaling in distributed memory. If you try to submit a run too big (either for RAM or scalability), Cloudrun will not let you proceed.

Any thoughts of supporting MPAS?

It's possible, we had a few people ask for MPAS already. Beyond WRF, the priority will probably be on surface wave modeling, but ultimately prospective users will decide.

Do you have preference for MPAS Atmosphere, Ocean, or both?

Surface wave, as in Wave Watch III? If so how about CICE and or HYCOM too?

My preference would be for MPAS-A.

Thanks

At the moment, I can't say much about the wave model that we will offer, other than that it will support both global and any custom regions, and it will not be WW3. NOAA already does as good job as you can get with WW3.

credit card form is broken. chrome and FF.

Doh! Thanks for the head's-up!

Reply on news.ycombinator.com