EDPS: Transitioning From 5G To 6G

At the recent EDPS, the keynote on the second day was by Mallik Tatipamula. He is the CTO of Ericsson Silicon Valley and also has experience all over the telecom industry with stints at F5, Juniper Networks, Cisco, Motorola, Nortel, and the Indian Institute of Technology Madras.

Mallik started with a potted history of mobile. You may already know all this, and I’ve covered my version of the story at length in:

• 1G Mobile: AMPS, TOPS, C-450, Radiocom 2000, and All Those Japanese Ones
• 2G: Mobile Goes Digital
• 3G and 4G: The Internet Arrives
• 5G: Connecting All the Things

1G was FDMA, frequency division multiple access. This means that each call was given its own frequency band. The radio technology was all analog. These networks were purely voice, and the phones would not work in different countries.

2G was TDMA, time division multiple access. This is a bit of a misnomer since it also used FDMA, and then divided the frequency band up into time slots. The technology was digital, with data rates of 16Kbps (for voice encoding). Very limited data came along. In the U.S., there was also one CDMA network, which stands for code division multiple access. This uses frequency hopping and constant updating of power so that all signals share the same spectrum. So these were voice and text messages, and with GSM (and, to a lesser extent IS-136) there was some interoperability of handsets in different countries.

3G was circuit-switched voice and packet-switched data at 150Kb/s. This was the first era with reasonable (at least by the standards of the day) mobile internet.

4G was true all-IP networking (voice just transmitted as voice-over-IP) with a performance at megabits per second. Finally, you could watch video on your phone in reasonable quality. Everything ran over internet protocol using OFDMA, or orthogonal frequency-division multiple access. This was the era of apps on your smartphone, the era of Uber, food delivery, chat programs, and so on.

In the early 1990s, GSM statistics were that there were 10M mobile subscribers around the world. In 2020, that 10M has grown to 7.5B. That has changed the lifestyles of everyone. The U.S. is actually not a leader in mobile commerce. In countries like China and India, vegetable sellers at the side of the road take (and expect) mobile payments.

We are now moving to 5G with beamforming antennas, and a lot of the economy is moving into the cloud. We may have gigabit-per-second speeds and low latency. Latency is a challenge since it is 50ms locally or 1s East Coast to West Coast. This means we need to process the data nearer the user, potentially right on top of the basestation. However, I think there is very little deployment of local data processing like that.

6G will come in the 2030 timeframe, meaning that the technical work is going on now. The expectation is very high bandwidth with both terrestrial (cellular and Wi-Fi) and non-terrestrial (space) integration. The spectrum bands are moving to 100s of GHz to low terahertz, which overlaps with radio sensing bands. In fact, we have some of those issues today with 5G phones interfering with airport navigation equipment.

We are moving towards convergence of connectivity, computing, control, and content.

Or, as Mallik put it more punchily, 3C + 3D = 6G. This is all coming together to satisfy low latency.

One thing I find a little odd is that mobile industry professionals always like to point out that low latency is a requirement for autonomous or assisted driving, with a car at 60mph going 17 feet in 200ms, so we need to get that down. The basic assertion is that 5G is essential for this. However, when automotive people speak, they never depend on network connectivity as part of automating driving. Sure, uploading maps and routing information, but cars have to be autonomous even in the presence of network failure, so they do not require 5G.

As with network infrastructure in the cloud/datacenter world, hardware is getting split from software through things like software-defined networking (SDN) and network function virtualization (NFV). This allows hardware to be replaced with virtual machines and having network functionality for the distributed era, as in the diagram above.

The next ten years compared to the last ten years is the story of moving from central cloud to distributed cloud. Or, as Mallik put it on one of his last slides, we are moving into a cyber-physical continuum. I have to confess, I don’t entirely know what this means. Yes, I know, lots of sensors, actuators, and electronics. But it sounds a bit too like ‘The Terminator’ being a “cybernetic organism.” We’ll just have to wait until 2030 to find what the future really brings.