Here is what you can obtain today by purchasing a generic Intel PC with a Software Defined Radio (SDR):
The total cost of this hardware is under 5000€, including the 20W digital amplifier (also known as Remote Radio Head) and antennae that are required to cover a radius of 20 km. The cost of Amarisoft software license required for the LTE is under 5000€ also. The cost of SlapOS and re6st software license required for the Edge Cloud, mesh backhaul and billing is zero since both are Free Software.
In other words, anyone today with sufficient skills could become a 5G telecommunication provider with an initial investment comparable to the cost of an entry level car.
The key component that has made this possible is Amarisoft LTE stack, authored by Fabrice Bellard, who previously created ffmpeg, qemu, linuxjs, etc. Even if few of us know it, we actually all use Fabrice's software inside our Android phones (ffmeg video processing) or each time we use services hosted by Cloud Computing providers.
Amarisoft LTE stack covers both the server side and the client side of LTE. The server side of LTE has two parts: the eNodeB and the core network. Both are supported by Amarisoft using a single binary code. Amarisoft eNodeB fully implements the concept of Virtual Radio Access Network (VRAN) by visualizing everything including the physical (PHY) layer. It is thus completely different from what some people call "small cell" in the sense that it is not a kind of downgraded and unreliable LTE stack. Amarisoft provides a full LTE with many features not even available in eNodeB products of Nokia, Ericsson, Huawei, ZTE, etc.
VRAN with Amarisoft LTE can be used to create autonomous cells, something that is highly desired in rural access or defense. Air-Lynx company has for example specialized in providing private LTE networks to police, firemen or military with a rugged platform that can be instantly deployed in the middle of nowhere.
VRAN with Amarisoft LTE can also be used to create high performance networks in cities, by combining dozens of Remote Radio Head (RRH), each of which with many antenna, and realize today the equivalent of massive MIMO by delivering 1 Gbps Internet to smartphones powered by Snapdragon 835 platform. The only hardware needed for this consists of standard RRH (3500€ each) connected to standard blade server (Advantech, Supermicro, OCP, etc.).
Amarisoft LTE software also implements LTE protocols for terminals. The Amarisoft UE software can be used to turn a standard PC into an LTE smartphone or into a high performance modem. Besides its obvious applications for testing or monitoring, it can be used to create network backhauls based on LTE protocol and thus the eliminates the need for a dedicated microwave infrastructure.
All provisioning, monitoring and deployment of services as well as backhaul routing that are required to setup a complete LTE 4G/5G network can be implemented with Nexedi's SlapOS. SlapOS is an open source edge computing operating system created in 2009 by Nexedi that happens to also be the only open source cloud solution that actually works for big data and mission critical systems. It can save at least 2 years to achieve integration of a complete VRAN / Edge Computing solution.
Thanks to SlapOS, the following features can be directly supported at the eNodeB level:
Because SlapOS does not rely on virtualization or on normal containers, it can optimise hardware usage and take into account execution constraints imposed by Amarisoft LTE stack.
Here are some results of a proof of concept that was deployed by combining Amarisoft LTE 100, SlapOS CDN and Eutelsat KASAT Internet backhaul. The goal was to prove that it is possible to deploy in half a day an LTE network in the country side with better performance than in Paris.
The hardware consists of a PC with an SDR card (left) and a KASAT modem (right):
And an antenna, mounted on a portable beach umbrella pole:
We can now test the performance of the network without cache: 22 Mbps download / 6 Mbps download / 700 ms latency.
Here are the results in the center of Paris: 6 Mbps download / 1 Mbps upload / 67 ms latency.
So, we are already faster than in the center Paris but with high latency. We tried VOIP with Facebook messenger and Skype. It worked well despite the 700 ms latency.
We then added SlapOS CDN system to handle our https sites. We could reduce latency to less than 30 ms. This means that this architecture could be even suitable for 3D online gaming or autonomous vehicle applications (less than 50 ms is required).
We heard that the KASAT link could be pushed to 100 Mbps if really needed. The cost of KASAT is not that high (less than 2€ / GB) and is comparable to the cost of LTE services in France.
In the future, we expect to setup a mesh and connect to another site by using an LTE hop. We could also connect to a local fiber and let our latency optimization routing protocol find out automatically the fastest route. Overall, it seems quite easy to provide shared 100 Mbps to about any village in Europe by combining KASAT and mesh networking over air or fiber.
The total cost per village of a complete system could be as low as 10 K€ (maybe even less). Providing broadband to the 70% of areas in France without acceptable network (should Paris be included in this definition ?) would probably cost less than 200 M€ with this technology.
Let us now review two cases where the combination of Nexedi SlapOS and Amarisoft LTE stack could lead to sustainable business.
In many developed countries, only 30% of the area of the country has access to LTE. Telecommunication operators are not investing in 70% of the country's area because they have no idea which area is going to be profitable. Also, with a cost of more than 100 to 150 K€ per site (instead of 10 K€ to 20 K€ for Amarisoft / SlapOS), traditional LTE infrastructure is not sustainable from an economic point of view.
By reselling their frequencies for limited duration (ex. 5 years) on those areas which they are not covering, Telecommunication operators could attract individual investors who would then deploy local infrastructure based on Amarisoft LTE stack and SlapOS Edge Computing. Also, by reselling their frequencies and delegating management of the infrastructure, telecommunication operators can save long and expensive negotiations with Huawei, Nokia, Ericsson or ZTE to interconnect Amarisoft LTE stack to traditional core networks (it works but maintenance contracts create a kind of non tariff barrier).
If operators are not willing to follow this approach, regulation authorities could actually do it instead so that no area remains uncovered. It is now proven that nation wide frequency licensing with obligations based on population ratio is not sufficient to deploy a country's broadband radio network. A local and automated approach for frequency licensing would however do the job much better.
More details about this business case are provided in Nexedi's reply to ARCEP (France's regulation agency for telecommunications): https://www.nexedi.com/NXD-ARCEP.RIP.PDF?format=
By attaching Amarisoft LTE stack to a cluster of inexpensive computers (ex. used OCP from Horizon Computing, rugged blades from Advantech), it is possible to cover a mid size city with a huge macro-station able to drive one hundred antennas at the same time, by combining multiple RRH configured in 4x4 MIMO and multiple frequency bands. The cost of this type of infrastructure is less than 100 K€ and can provide 1 Gbps LTE to all inhabitants. It is even possible to deploy on the inexpensive cluster a low latency 3G gaming solution such as Shadow.
Here is a list of suppliers to take into account besides those listed previously in the article: