Trillion Blue Oceans: Deterministic Networks and the Metaverse
Author| Huang Yudong, Ph.1 student at the State Key Laboratory of Networks and Switching, Beijing University of Posts and Telecommunications. His research interests include network architecture and deterministic networks. Email address: hyduni@163.com.
At present, deterministic networks have become a hot and urgently needed focus direction of global research. On the one hand, in the context of the grand strategy of the Industrial Internet, industries such as steel, petroleum, mines, ports, logistics, and manufacturing are deeply integrated with ICT technologies, and access devices have grown in large numbers. Traditional bus networks are gradually being standardized by deterministic Ethernet. Instead, the deterministic network will help the industry to transform and upgrade to intelligent and digital, and realize the improvement of quality, cost reduction and efficiency. On the other hand, emerging time-sensitive scenarios and applications, such as car networking, metaverse, holographic communication, remote teleoperation, remote diagnosis and treatment, remote driving, remote equipment maintenance, immersive real-time tourism, interactive distance education, and online interaction Playing concerts and tactile Internet have stimulated people's greater imagination for the future Internet. A network with strictly guaranteed service quality such as bandwidth, delay, jitter, and packet loss will open a new era of interconnection and bring A trillion-level blue ocean market.
Taking the application of the Metaverse as an example, from the perspective of technology development history, the Internet is evolving from an app application that transmits information to a Metaverse that is deeply integrated with production, life, health, and entertainment. From 2010 to 2020, with the upgrade of mobile Internet, cloud computing and other infrastructure, national-level applications such as Douyin, WeChat, and Taobao were born. Since 2020, a new round of industrial technological transformation has begun. Facebook has changed its name to Meta, Google Map has launched an immersive real-life 3D map, Microsoft has acquired game giant Activision Blizzard, Byte has acquired VR startup Pico, and Tencent has launched XR full-real Internet. Domestic and foreign giants have laid out to seize the commanding heights of the "metaverse". In addition, for hardware manufacturers such as Apple and Xiaomi, the global mobile phone shipments fell by 11% in the first quarter of 2022, and the domestic mobile phone market shipments in March fell by 40.5% year-on-year. Those headsets that can bring a new interactive experience , glasses, earphones, watches, bracelets, handles and other smart wearable devices may subvert the mobile phone screen and become a new growth point.
In the next ten years, whether the network infrastructure can realize massive "human-machine-object" real-time interconnection and interoperability will be the key to whether China can take the lead in nurturing the metaverse. However, the Internet, as the core link, does not currently have the ability to guarantee service quality such as deterministic delay, jitter, and packet loss rate. Uncertain delay will lead to immersive experience freezes, interactive dizziness, and user persuasion. , and even safety accidents may occur. Therefore, the application must come before the network.
The network itself is a complex system full of uncertainty, which includes terminal equipment, network equipment, and computing and storage resources; horizontally, the network must pass through multiple autonomous domains, such as local area networks, access networks, aggregation networks, backbone networks, Cloud data center; Vertically, the network covers 5G, WiFi, optical, Ethernet, TCP/IP, Quic, http and other sophisticated protocol technologies at different levels from hardware boards to software operating systems. The Internet has been tempered for more than 50 years. "Do your best" forwarding is its foundation. To require it to be certain is a bit square and round, like new shoes that look good and want to buy always don't fit well.
In order to allow the traditional network to smoothly evolve into a deterministic network, and to facilitate discussion and research, this paper proposes three possible principles for everyone to define a deterministic network.
Deterministic networking is premised on deterministic traffic
If we can strictly guarantee the quality of service of the application, it must be because we know enough about the traffic of the application. Computers and mobile phones generate user traffic. Generally, the traffic arrival curve can be regarded as a Poisson distribution, that is, when the user is using the traffic, the traffic continues for a period of time, and becomes very small when not in use. When the massive traffic converges, the curve becomes Normal distribution, so it has long-tail delay; traditional networks are not aware of applications, and can only try their best to optimize bandwidth and reduce average delay. In industrial scenarios, sensors, controllers, and actuators generate machine traffic, and their traffic characteristics are very clear. For example, a control command sends a 100-byte data packet every 1ms, occupying 0.8Mbps of bandwidth, and its arrival curve is an even distribution without bursts.
In addition, the characteristics of slow start and fast retransmission of computer TCP congestion control will bring a large number of traffic bursts, and many industrial equipment (such as PLC) do not have TCP/IP protocol stack, which is directly encapsulated into a frame in the second-layer MAC and sent. out. Taking the in-vehicle network as an example, in addition to audio-visual entertainment, most of them are deterministic traffic of machines, which conforms to this principle. Therefore, the time-sensitive network technology based on Ethernet is used to replace the chaotic bus in the car, so as to reduce the body weight. , It is completely feasible to increase the body space and improve the intelligent level of the car.
Then if the traffic is a little more complicated and not completely deterministic, can a deterministic network still work? can. At this time, we need to let the network perceive the application, sign an SLA service level agreement with the application, such as the commitment to the maximum bandwidth and burst rate, and then reserve network resources such as bandwidth, time slots, and cache along the path according to requirements. Therefore, deterministic networks are usually also related to technologies such as global control, access control, connection establishment, resource reservation, and network calculus. If the traffic does not come as promised, the over-delivered traffic may be discarded, or various traffic shaping methods are used to make it meet the promised traffic arrival model. Of course, the shaping itself will introduce additional delay. Furthermore, if it is completely arbitrary traffic, then there is no way to talk about a deterministic network.
Therefore, "determinism" is equivalent to adding constraints. The more constraints, the more certain it is. "Deterministic flow" is the biggest prerequisite constraint in "deterministic network".
Deterministic traffic is only a small fraction
Suppose we build a deterministic network and the port rate is 100G, can we carry 100G deterministic traffic? It's almost impossible. However, if the deterministic traffic only accounts for 10G, and the remaining 90G is still best-effort traffic, this is more feasible. Next, we use the canteen queuing scene and the ambulance scene to analyze why the deterministic flow is best only a small part.
In the canteen queuing scenario, suppose a company has 1,000 employees who all go to the canteen to eat at the same time after get off work at 12:00 noon. Those at the back of the line will inevitably face longer waiting times. Since everyone is equal, no matter how the scheduling is optimized or how the queue is cut, the last "deterministic employee" cannot get a good guarantee of delay and jitter. Network resources are limited, and the certainty of the network and the reusability of resources are the two poles of a magnet. For example, a dedicated line that pulls optical fibers is highly deterministic, but it can only transmit point-to-point, without routing switching and multicasting, and the resource reuse rate is low; Ethernet uses statistical multiplexing, and the resource utilization rate is high, but when the bandwidth utilization rate is When it reaches 30%-50% or higher, a large amount of packet loss will occur, and hard slicing, priority queues and other methods are required for isolation, which reduces the resource reuse rate and improves the deterministic guarantee.
In the ambulance scenario, there is a dedicated emergency lane on the highway, and ordinary vehicles will also take the initiative to give up road resources for the ambulance, so the ambulance has the right to prioritize transmission and can reach the destination within a certain delay. In the case of many ambulances, we can also adjust the traffic lights and other methods to keep the ambulances at a distance from each other and isolate them in time and space. In addition, this is also a good explanation for the deterministic network will not replace the traditional Internet, because it is forward compatible, traditional best-effort traffic can still run on ordinary roads.
Therefore, certainty is equivalent to priority, and priority is destined to be only a small part of certainty.
Determinism, flexibility, and scalability can only take two
In the early technological evolution, there was a dispute between ATM and Ethernet. ATM can strictly guarantee the quality of service by establishing a connection to transmit cells in a similar circuit-switched manner, but it has not been widely used. At the same time, there are two models of DiffServ and IntServ in the QoS technology of the Internet. IntServ can also provide deterministic service quality assurance for special applications by means of end-to-end RSVP resource reservation. However, due to the need to maintain the state of each flow, the lack of technology Scalability, unable to cope with massive traffic scenarios, so it has not been widely used. Ultimately, the flexibility and scalability of Ethernet/IP will rule the world.
Because Ethernet has the advantages of large bandwidth, low cost, and good compatibility, now we want to implement a deterministic network based on Ethernet, and we have to give up. If certainty is required, flexibility and scalability cannot have both. For example, within the scope of a local area network, we can build a dynamic and flexible deterministic network through some technologies, but it will be limited by factors such as network radius and hop count. If we want to expand to a large network, and want to achieve deterministic transmission in complex topologies and massive traffic scenarios, if we want scalability, the flexibility will be greatly reduced. If a solution has both certainty, flexibility, and scalability, then we probably need to consider whether its technical complexity and cost are appropriate, as well as the possibility of commercial landing.