Bandwidth is often referred to as speed, but it actually refers to how much data can be transmitted in a given time. The requirements for the Metaverse require more bandwidth than many internet games and applications, and are beyond most modern connections. This is best understood using Microsoft Flight Simulator.
Microsoft Flight Simulatoris a most accurate and extensive consumer simulation. It contains 1.5 billion buildings, 2 trillion trees individually rendered, and almost every road, mountain and city worldwide. All of these look exactly like the real thing, as they are based on high quality scans of the actual thing. Microsoft Flight Simulator requires more than 2.5 petabytes (or 2,500,000 gigabytes) of data to achieve this feat. This amount of data cannot be stored on a consumer device or most enterprise devices.
Microsoft Flight Simulatoris an live service which updates to reflect real world weather conditions (including wind speed, direction, temperature and humidity as well as lighting and traffic). It is possible to fly directly into real-world storms while following IRL commercial aircraftliners along their exact flight path.
Microsoft Flight Simulatorworks in that it stores a core amount on your device. This data is used to run the game as well, just like console games and not cloud-based streaming services like Stadia. However, when players are online, Microsoft streams huge amounts of data to their local device as needed. It's like a pilot flying in real life. They see new information when they cross a mountain or go around a bend. This helps them to clarify and reveal what is there. They have no knowledge but will give them.
This is the assumption of many gamers who believe that this is what happens to all online multiplayer games. In reality, game services only send positional data and player input data (e.g. Shoot, throw bomb), or summary-level data (e.g. Players remaining in a battle royale). To individual players. You already have all of the rendering and asset data on your device so there is no need to download or install it.
Games can offer a greater variety of assets, items and environments by sending rendering data as needed. They can do this without the need for lengthy downloads, installations, update batching or huge user hard drives. Many games now embrace this hybrid model, which combines data streaming and locally stored information. This approach is especially important for Metaverse-focused platforms. Roblox is an example. It requires (and benefits more) from asset, item and environmental diversity than titles like Mario Kart or Call of Duty.
The complexity and importance of virtual sim will grow, so the data that must be streamed will also increase. Robloxreceives benefits from the fact that many of the underlying assets and prefabs can be reused and modified in a light manner. Given this, Roblox is mostly streaming data on how to tweak previously-downloaded items. The virtual platform will eventually need to be able predict the results of nearly all possible combinations and creations.
Virtual twinning platforms, also known as "mirrorworlds", such as Microsoft Flight Simulator already have to reproduce the almost infinite (and easily proven) diversity of real life. This means that you can send far more data (i.e. This means that you will send far more data (i.e.,. It's exactly the same dark cloud as this. This data is constantly changing, which is crucial.
This is the most important point. This is crucial if we are to be able to interact in a vast, real-time, shared and persistent virtual environment.
The'real world' can be compared to Fortnite’s map. Everybody on Earth is in the same simulation, simultaneously, and with complete permanence. A tree that I have cut down is gone forever. You can only access Fortnite via a fixed, current version of the map. Everything you do in that map can only be shared with a few users and it will reset after a brief period. How can you cut down a tree? It will be reset in 1-25 minutes. The map was never accessible to more than 99 users. Epic Games only sends a new version of the map. Epic Games could send your universe out to everyone. This is fine for many virtual experiences. This will also work for Metaverse-specific experiences. Some experiences, if not all of them, will require persistence from all users at all times.
If we want to seamlessly move between virtual worlds, cloud data streaming is essential. Fortnite's Travis Scott concert featured seamlessly transporting players from the game’s core map to the depths a never before-seen ocean, then to an never-beforeseen planet, then into outer space. Epic sent these worlds to all users via a standard Fortnitepatch days before the event. This meant that users wouldn't have the ability to play the event if they hadn't downloaded the update. During each set piece, players' devices were loading the next setpiece in the background. Although this system is extremely effective, it requires that a publisher knows which worlds users will visit next and how long ahead. You can choose to download all possible options, which is impossible, or to cloud stream them.
There is also incremental player data, in addition to the increased environmental data. The Fortnite server will send you information about your friend and their activities. You don't need to do anything except reload an assault rifle or fall. This information must be sent to your friend if you have real-time motion capture. As well as everyone else. You can also stream a video file from this game, Fortnite occasionally offers. You want to hear the spatial audio from a crowd? The same. Have you ever felt a stranger touch your shoulder with a haptic bodysuit. The same.
Online games that only require input and position data have many players experiencing bandwidth issues and network congestion. These needs will only be exacerbated by the Metaverse. The good news is that worldwide broadband penetration is increasing and bandwidth is also improving. Compute, which is discussed in Section #3, can also improve and substitute for constrained data transmission. It can predict what will occur, until the point when the'real data' can be substituted in.
Latency
Latency is the biggest problem in networking, but it's also the most difficult to understand. Latency is the amount of time data takes to travel from one place to another. Latency is often regarded as the least important KPI, in comparison to reliability (above and below) or network bandwidth (below). Because most internet traffic is either one-way or sync, this is a common KPI. It doesn't matter how long it takes to send a WhatsApp message or receive a read receipt. It doesn't matter how long it takes to stop YouTube's video. It's important to watch Netflix continuously, not just right away. Netflix will delay the start of a stream to make sure your device is able to download the video stream ahead of you are watching it. This way, if your network has a problem, it won't be noticeable.
Video calls that are both synchronous and persistent have a high tolerance for latency. Video is considered the least important element of a call, so video-calling software will prioritize audio if there is a network problem. Software can help you reduce latency by increasing the playback speed and quickly editing out pauses. It's also easy for participants manage latency. All they have to do is wait a little.
Low latency is required for the most immersive AAA online multiplayer games. Latency is the time taken for information to be received by a player (e.g. Latency is the ability to determine where a player is and whether a grenade or soccer ball has been thrown. It also affects how fast information is relayed to other players. In other words, latency determines whether you win, lose, kill, or get killed. Modern games run at between 2 and 4 times the average frame rate of video. This is why many people have adopted these higher frame rates, even though they are not as popular with traditional video. It is essential for performance.
Video gaming is far less likely to cause latency than other media. Take, for instance, traditional video and video games. Unless audio arrives more than 45ms early or 125ms late (a total of 170ms), the average person won't notice that it isn't in sync with video. The acceptable thresholds for acceptance are higher at 90ms (275ms) and 185ms (275ms). Digital buttons such as the YouTube pause button are digital buttons. We only consider our clicks to be unsuccessful if we don’t get a response within 200-250ms. AAA games are frustrating for avid gamers at 50ms, and non-gamers at 110ms. Games become unplayable after 150ms. Subspace On average, a 10ms decrease or increase in latency results in a 6% reduction or an increase of weekly play time. This is a unique exposure that no other business has.
Let's now look at the average latency worldwide, keeping in mind the bands. The median roundtrip time to send data from one place to another in the United States is 35ms. This is not the case for all pairings, particularly in dense cities or areas with high demand peaks (e.g. San Francisco to New York in the evening. There's also the transit time from one city to another, which can be prone to slowdowns. Congested areas, such as dense cities or neighborhoods, can lead to congestion. You can also play via 4G technology, which averages 40ms. If you don't live in a major city, your data might have to travel 100 miles on an outdated, poorly maintained wireline infrastructure. The median delivery delay between cities is 100-200ms globally.
The online gaming industry has developed several partial solutions and hacks to manage latency. None of these solutions scale well.
High-fidelity multiplayer gaming, for example, is often'match made around' server regions. Publishers can reduce latency by limiting the player list to those who reside in the Northeast United States, Western Europe, and Southeast Asia. This clustering is good enough since gaming is a recreational activity that's usually played with one to three friends. It's unlikely that you will play with someone who is located several time zones away. You don't care about where your unknown opponents live (who you rarely even have the chance to talk to). Subspace still finds that around three quarters of internet connections in the Middle East have latency levels beyond what is necessary to play dynamic multiplayer games. This contrasts with the quarter of those in the United States or Europe. This is mostly due to the limitations of broadband infrastructure and not server placement.
Multiplayer online games use netcode' solutions in order to maintain synchronization and consistency. A delay-based netcode tells a player's device (e.g. A PlayStation 5 will artificially delay the device's rendering of inputs from its owner until they are received by the more latent player (i.e. Their opponent's inputs arrive. It is annoying for players who have muscle memory that is sensitive to low latency but it works. Rollback netcode is much more complex. A player's device will follow the expected path if an opponent's inputs are delayed. If the opponent does something unexpected, the device will attempt to unwind the in-process animations and replay them "correctly".
These solutions are great for 1v1 (e.g. 2D fighters, for minor latency hiccups. +-40ms) and titles that have a restricted range of highly predictable actions. A driving game, or a 2D fighter. These solutions become less effective as we move to Metaverse-focused experiences, with more players, more latency variations, and more dynamic situations. It is difficult to predict and correctly predict 12 players and then 'roll them back’ in a non-disruptive manner. It is more sensible to just disconnect a laggy participant. Even though a video conference has many participants it only matters to one person at a given time. There is therefore a core latency. Latency increases when all players are able to get the correct information.
Most games don't have high latency. Titles like Hearthstone or Words With Friends can be turn-based, synchronous, or both. Other hits like Honour of Kings and Candy Crush don't require pixel-perfect inputs. Low latency is only required for fast-twitch titles such as Fortnite and Call of Duty. These games can be lucrative but only a small percentage of the total games market, based on titles produced -- and even less of total game time.
The Metaverse is not a fast-twitch AAA video game, but its social nature and desired importance mean it will need low latency. Human conversation is greatly influenced by facial movements. We are also sensitive to small mistakes and issues with synchronization (hence, the uncanny valley issue in CGI). Their ubiquity is also crucial for social products. Imagine FaceTime and Facebook not working if your family or friends were located within 500 miles of you. Or only when you were home. We need more bandwidth than that if we want access to remote or foreign labor in the virtual universe.
Latency is among the most difficult and slowest network attributes to resolve. As mentioned, the problem stems partly from how few applications and services require ultra-low latency delivery. This limits the business case of any network operator or latency-focused Content-Delivery Network (CDN), and is in conflict with the fundamental laws and physics.
It takes between 40 and 45 minutes to travel from NYC or Tokyo to Mumbai at 11,000-12,500km. This meets all low latency thresholds. Fiber optics makes up the majority of internet backbone. However, fiber-optic cables fall 30% below the speed of light because it is rarely in a vacuum (+ loss of 3.5 dB/km). Copper and coaxial cables suffer from even greater latency degradation over distance and have a lower bandwidth. This means that they are more susceptible to congestion and delays. These cables are still a large part of the ones found in interiors of residential and commercial buildings, as well as in neighborhoods.
These cables are not laid in the same way as the crow flies. What we think of as the internet backbone is actually a loose federation private networks that do not fully deliver data packets. The network distance between servers or clients can be significantly greater than their geographical distance. Network congestion can cause traffic to be routed less directly to ensure reliable delivery and continuous delivery. The average latency between NYC and Tokyo is more than 4x, while NYC to Mumbai takes 4-6x.
Relaying any cable-based infrastructure is expensive and time-consuming, especially if you are trying to reduce geographic distance. It also requires considerable regulatory/government approval, typically at many levels. Wireless is easier to fix. 5G is certainly a help, with its average 20-40ms reduction in 4G (and promise as low as 1ms latency). This only improves data transmission over the last few hundred metres. After your data reaches the tower, it returns to traditional backbones.
Starlink, SpaceX’s satellite Internet constellation company promises high-bandwidth and low-latency internet service throughout the United States and eventually around the globe. However, this does not solve the problem of ultra-low latency at long distances. Starlink can travel between your home and the satellite in 18 to 35ms. However, this time extends to when data needs to go from New York City to Los Angeles. This requires data to be relayed across multiple satellites. In some cases, Starlink even exacerbates travel distances. New York to Philadelphia travels approximately 100 miles straight-line and potentially 125 by cable. However, it can travel over 700 miles to reach a low orbit satellite and back down. Fiber-optic cables are also less susceptible to loss than light transmitted through the atmosphere, particularly on cloudy days. Also, dense urban areas can be noisy and susceptible to interference. Elon Musk stated that Starlink was focused on "the hardest-to-serve customer that [telecommunications company] otherwise have difficulty reaching" in 2020. In this sense, Starlink brings more into the Metaverse than it boosts those already involved.
To meet the increasing demand for real-time bandwidth applications, entirely new technologies, business lines and services are being created. Subspace (Disclosure : Portfolio company) deploys hardware in hundreds of cities to create 'weather maps' that will help to find low latency network paths. It then operates a network stack that coordinates the needs for a low latency app with all the third-parties along this path. The optical network can also be spliced across multiple fiber networks to reduce the distance between servers and minimize non-fiber cabling.
Fastly provides a CDN that is optimized for low latency applications and not just bandwidth or delivery reliability. The company uses an 'infrastructure-as-code' approach that allows clients to customize nearly every aspect of the company's edge-computing clusters, promises that a software application can clear and replace all cached content across all of these clusters globally within 150ms, and that it can cache and accelerate individual blockchain transactions in real time.
Reliability
It is quite obvious that reliability is important. Reliable quality of service is a key factor in our ability to transition to virtual labor and education. This includes both the overall uptime and consistency of other attributes, such as bandwidth, latency, and download/upload bandwidth. Many people who "live online" today might find this alarming. Netflix streams in 1080p and even 4K perfectly well most of the times! Netflix, however, leverages reliability solutions that aren't compatible with games or Metaverse-specific apps.
Netflix and other non-live video services receive video files for hours or months before they are made available to the public. They can then perform extensive analysis to reduce (or 'compress) file sizes. This includes analyzing frame data to determine which information can be deleted. Netflix's algorithms can 'watch' a scene that has blue skies and determine that if the viewer's internet speed drops to 500, 50 or 25, then 500 different shades of blue could be reduced to 200. Analytics on Netflix even recognize that scenes with dialogue may be more compressible than scenes with fast-paced action. Multipass encoding is used here. Netflix uses the spare bandwidth to send video to users' devices before they are needed. This means that even if there's a temporary drop or increase in latency, no changes to the user's experience. Netflix also preloads content from local nodes so that when you request the latest episode of Stranger Things it is only a few blocks away. This is not possible for live video or data, which must arrive quicker. It's therefore more difficult to cloud-stream 1GB or more of Stadia than 1GB Netflix.
So even though its objective isn't necessarily competitive in nature, we should think of the Metaverse as raising the requirements for all aspects of networking -- latency, reliability/resilience, and bandwidth -- to that of AAA multiplayer games. It doesn't matter what kind of device you have (see hardware, compute), if your device can't get all the information it requires in a timely manner it's not worth it.
Source: https://www.matthewball.vc/all/networkingmetaverse
Comments
Post a Comment