I believe the biggest argument here is the sheer demand to reduce trade times by 5ms or so
between their points of presence.
I haven’t seen the REAL engineering, but I have a lot of experience with modeling and optimizing capacities of modern networking gear. The proposal seems to claim a near zero latency of the data stream from a computational decision that
has an “advantage expiration time” under 5ms.
In 2017, I did a series of simulations for Magic Leap
https://www.magicleap.com/en-us/ to predict their user experience based on Tprop and expected latency.
We purchased several 10Gb “dedicated connections” from the Miami data center and other major data centers – LA, Phoenix, Chicago, NYC and Atlanta were among the cities.
After passing through a few wire runs and switch ports, these guys will be lucky to see 3ms round trip latency to their transmitter in the same building.
That’s all we could get out of the building using memory resident transfers, bonded 10G connections to the newest QFX10000 Juniper switches.
So they’re going to be able to compress, RF modulate and put it on the wire in less than 3 ms? Doubtful. They’re going to use something that’s not TCP/IP?
If they wait – maybe 5, certainly 10 – zero transfer rate quantum networking will be commercially available. It is being used right now for military and robust commercial real-time sharing of encryption keys.
This reduces the site to site Tprop to near zero.
Quantum networking systems are in customer testing scenarios right now, primarily for cryptographic reasons and because data transfer rates are not high. The transfer rate is currently the target of research and schemes to make it cost
effective are underway.
BUT – when the management of broad area quantum networks becomes available commercially, the RF stuff will have little value.
--
Mickey Baker, N4MB
Director, Southeastern Division
ARRL,
The National Association for Amateur Radio®
Phone (561) 320-2775
Email:
n4mb@arrl.org
ARRL Customer Service:
(860) 594-0355