Practical long distance networking (>500km)

Fiber can easily reach your 100G target.

Non-Fiber options will never get close... not at least a 100G speed, this is just to fast.

Thank you David for your comment... I am willing to create high quantities of non-Fiber options, because in my position fiber-optics if not an option at all... so even the next best thing would be more practical.

Any ideas?

Laser is not reliable in the atmosphere at that distance.

The unlicensed radio bands can't punch a signal that far and commercial operation is prohibited in amateur bands.  That means a commercial radio license is required and commercial-grade type-approved radio equipment.

The only bands capable of handling that data rate are well up into the microwave region and microwaves are strongly attenuated by atmospheric water.  Terrestrial microwave links were used up through the 1960s but were dumped as fiber came along and that is due to the many problems associated with the medium.

https://en.wikipedia.org/wiki/Microwave_transmission

"... The longest microwave radio relay known up to date crosses the Red Sea with a 360 km (200 mi) hop ..."

So laser won't do and terrestrial microwave won't do, and those being the only media that can handle the necessary bandwidth, I'm forced to say that the project is not possible, well before even starting to look at pushing data at that speed through a very unreliable channel.

Hi Dr Klanh,
Thank you very much for your ideas and advise. However, in my case, fiber optics is just not possible at all... so I am willing to go commercial and acquire required licenses.
Any ideas or insight?
Thanks.

This is something that requires the expertise of a company which does it for a living.  Wading through the details of obtaining a commercial license for data is (at least in the USA) not something that can be done without a company that has an EE who specializes in RF, another EE who specializes in data transmission, and a lawyer to deal with all the issues the government will throw in your way.

I would think that two million dollars would be enough to get through the planning stage and acquire the necessary licenses, and another two million for hardware, and then you can start looking at how to site the equipment.  So a very strong commitment is required to spend that kind of money.

ok but are there any other ready-made silutions done by other companies for this case that you may know of?
Thank you very much!

In my previous comment I was figuring on buying ready-made commercial equipment.  Type acceptance prohibits using anything else.

The consulting company EE who does this for a living has to look at the problem and say "Well, maybe it can be done ... with these constraints" and that's where the first two million bucks is going even before thinking about buying hardware.

https://www.everythingrf.com/rf-calculators/line-of-sight-calculator  says for 500km hop you would need an antenna height of 65,500ft. I vote for fiber. 

500Km with a sea separation in my mind means from one country to another. This means dealing with 2 government bodies .  If latency isn't an issue then satellite is your only feasible option.
You could use the internet but it means a peering arrangement with 2 telecommunication companies that already have peering arrangements in place. 10 GiB for $1500 x 10 = 15,000/month (3 year contract) at both ends.

There's always satellite and the necessary speed can probably be obtained by leasing a private transponder, but the 1/4 second latency is probably unacceptable.

https://www.satsig.net/ivsatcos.htm

"For Ka band HTS satellites, the cost per Mbit/s is now just under about $200 per Mbit/s per month !"

Let's put that into perspective against consumer terrestrial connection costs.  Assuming that price can be obtained the equivalent of a cable connection at 40 Mbit/sec would be $8,000 per month.  Obtaining 1 Gbit/sec would be $200,000 per month.  10 Gbit/sec $2M/month, and 100 Gbit/sec $20M/month.  So even if the 1/4 second latency is acceptable, the first year's service would be right around $240 million bucks, less any possible discount for purchasing bandwidth in bulk.