franksingleton
Well-Known Member
Better forecasts ahead as Met Office transitions to a supercomputer in Azure cloud describes the recent introduction of their latest computing facility and the reasons for its introduction.
Latest Met Office supercomputer
- Thread starter franksingleton
- Start date
fredrussell
Well-Known Member
Last edited:
newtothis
Well-Known Member
An upgrade from:
Snowgoose-1
Well-Known Member
Thanks for posting.
Hope it suits some folks.
I have given up on the met office sailing forecasts for some time now due to the wide banding of wind forces. This is frustrating as we all contribute money to this organisation whether we like it or not.
They have over 1400 people working at Exeter and over 2200 globally . You wonder what is going on.
I doubt if the super computer is going reduce staff members.
Sorry Frank . I know you mean well.
Hope it suits some folks.
I have given up on the met office sailing forecasts for some time now due to the wide banding of wind forces. This is frustrating as we all contribute money to this organisation whether we like it or not.
They have over 1400 people working at Exeter and over 2200 globally . You wonder what is going on.
I doubt if the super computer is going reduce staff members.
Sorry Frank . I know you mean well.
franksingleton
Well-Known Member
Never forget that the amount of time given to matine forecasts is determined by the NCA in consultation with users. For longer than I remember - a very long time - marine forecasts texts have always been akin to squeezing a quart into a pint pot~. The classical example is the BBC Shipping `forecast. A maximum of 330 words in total. NAVTEX slots are small. The official marine forecasts can never be more than headlines and warnings.
The forecasts that we see are a small part of what the Met Office does. Try asking Google to list Met Office functions.
I put up the OP because, as far as operational forecasting is concerned, we are at an interesting stage where AI with its quite different computing needs could well become an important tool. Climate studies and research are becoming/have become major issues for us all. It is not at all clear what the mix of AI and physical modelling will be.
Boathook
Well-Known Member
With all this super computer and AI, will they increase the inshore forecast to cover 3 days ?
franksingleton
Well-Known Member
Perhaps my #5 was not detailed enough. The Met Office will provide services as agreed with the MCA. I suggest that you contact the RYA. They have (or had in my time) a seat on a committee that represented all marine users. It is not for the Met Office to decide what forecasts are produced. They will advise or comment on requests and then provide an agreed service.
billskip
Well-Known Member
Wasn't that a total failure?
The government then paid millions for an upgrade and they turned it sideways.
lustyd
Well-Known Member
Weird that they didn’t seem to mention what it actually is. “Supercomputer” is a pretty outdated tech and approach and although you can buy that through Azure it offers essentially none of the benefits of Azure aside from collocated networking to other cloud services. They allude to this in the article, speaking of AI augmentation, but for real benefit they ought to look at reworking to a real cloud approach with small compute instances.
franksingleton
Well-Known Member
I suspect that, at this stage, they do not know what the mix will be between AI and physical modelling in both operational and climate prediction. There are also research projects into understanding the atmosphere - eg the Bayesian storm.
dolabriform
Well-Known Member
The biggest problem I see is that they have decided to partner with M$.
Late to school, copies everyone else's homework, and cuts corners on security and reliability.
Late to school, copies everyone else's homework, and cuts corners on security and reliability.
franksingleton
Well-Known Member
I think that you misunderstand the nature of the Met Office/MS relationship. The Met Office has long been in the forefront of using computing technology right from the time that they were using LEO, the Lyon’s Tea House computer, through the Aldermaston, GFDL (USA)/NASA) computers, the in house computers from the English Electric Mercury, various Ferranti, IBM, Crays etc. You might even go back to the Pilot balloon slide rule which had 2 cursors, invented in 1915. A skilled observer, single handedly, could follow a balloon with a manual theodolite while inputting azimuth and elevation into his slide rule and calculate winds during the ascent.
Now, the power and space requirements for the computing power they need is such that maintenance of an in-house facility is no longer realistic. My understanding is that MS will host the hardware and, no doubt, some of the communications. Increases in communications speeds and capacity were, no doubt, an important factor preventing earlier implementation. They have operational requirements to meet such as aviation and defence services.
Whether Azure is the best option, is way beyond me. The development of AI has added another dimension to the computing/communications requirements.
Last edited:
lustyd
Well-Known Member
MS have consistently been at the front with AI and ML workloads.
I won’t comment on the rest but they certainly weren’t late this time and neither did they copy anything. With ML and data science they’ve been very heavily copied, even the failed products were copied
franksingleton
Well-Known Member
Perhaps we are running out of eagle eyed cognoscenti. At #12, I should have said, Ferranti Mercury, English Electric KDF9……. Memory at fault. Not important, just noticed the errors.
