Putting a Price on Latency: Startup Network Next Aims to Fix Internet Routing
lightreading.com
lightreading.com
If you have 30 ms latency (e.g. my DSL to Riot Games) then you had better have the auction over and done with in much less than 30 ms.
Also the latency that people perceive is not the 30 ms average latency, it is the spikes of 300 ms or more latency that sometimes happen. I think what people hate about computers the most is the way they become unresponsive for a few moments and you don't know if it is going to come back OK or not.
I would rather have 30 ms latency that is rock solid as opposed to 20 ms average with 99% at 200 ms. Any route flapping at all is going to contribute to the latter.
https://www.polygon.com/2016/2/17/11037220/league-of-legends...
> it is the spikes of 300 ms or more latency that sometimes happen
yes that's jitter, that's part of what they're trying to address. That's like ... the whole point; you're paying to avoid routing through intermediaries with unreliable performance characteristics.
edit: I see your other comment now. It wasn't there when I wrote this comment. Leaving it as it was for posterity.
The thing about a lot of residential connections is that probably a not-insignificant amount of the problems are very close to last mile, as opposed to bandwidth availability at backbone or major interconnection points
Also many people seem to think it is ritually unclean to use wires so they use WiFi in cases where ethernet would perform much better.
From a business perspective, that is what makes it so hard. You could do the things and spend the money to get low latency at some steps in the chain and lose it all because of something that happens sporadically at another one.
That is why I Bring Riot up as an example because the "vertical integration" there reduces the number of parties would could be involved and makes it a little more possible that they could invest in facilities and have customers perceive the results.
Limiting it to low-bandwidth would discourage misuse and could enable offloading the decision of how and when to use it to the end user or application. Some users could reserve it for gaming, others for voip, others for something else. Some might try using it for netflix or web browsing but they would probably find that to be undesirable because of the low bandwidth.
You don't really need a private network for it either - there's no reason you couldn't have low-latency/low-bandwidth peering across regular ISPs.
The crux of this system is that the amount of benefit you get out of it is a function of how much money you put into it. I can't believe that he would say ... this, of all things:
> What happening that I think is terrible is that the public Internet is turning into this neglected commons
He's directly creating a platform that exacerbates that very problem by finding loopholes in how network neutrality policy is structured to create fast lanes. This in turn will create incentives for providers to put their most performant systems on private networks and their least performant systems on the public internet. The greater the gap in performance between the public internet and the services of Network Next, the more valuable Network Next is. It's a feedback loop that will print money. It's a brilliant act of capitalism that is doomed to succeed at the expense of the already-sparse indie multiplayer scene.
This is a past tense. It has already happened. What if indie games could access the same benefits of something like Riot Direct, democratized so they can use it at a reasonable price without having to build 10s of millions of dollars worth of their own private network infrastructure?
Where we're doing at Network Next is democratizing access to the private networks, not creating more of them.
It strikes me as odd to describe your actions as "democratizing access" when what you're doing is basically the textbook definition of market making. Markets structure access by capital; democracies are ideally structured to provide equal access that is unrelated to participants' access to capital. By transforming private networks into fungible commodities, you will ideally raise the utilization of said private networks and in turn raise the return on investment of building such networks.
You say you're not creating more private networks, but again, I think this is a useful turn of phrase that is not representative of the consequences of your actions. Even when you're not making more private networks yourself, if you build an efficient market for commodifying access to private networks, shouldn't that create incentives for others to create new private networks? I'm not questioning your motives; I'm questioning what the consequences of your actions will be.
Since the bill here is footed by the game developer (and not the player), the cost barrier that indies face in making multiplayer games is further tilted in the favor of AAA studios. In short: great for AAA studios since it makes it harder for smaller studios to compete, great for private network operators since it provides liquidity, great for gamers that want to play AAA multiplayer games, bad for indie game studios because it will widen the latency gap between public and private networks, bad for gamers that want to play indie multiplayer games.
The end result of this appears to be a system whose consequence will be to create market forces that encourage the centralization of capital and the creation of new private networks, while also adversely affecting the landscape of which game developers are capable of creating latency-sensitive multiplayer games. That is what I find concerning.
What we do is open up existing private networks and resell excess capacity on them. There's more than enough excess capacity on existing private networks to satisfy the needs of game traffic, and we create a neutral marketplace that opens this up so game traffic can go across these networks.
Think of this in terms of turning cloud backbones, CDN private networks into carriers of game traffic... nobody is going to build a private network for games, the traffic volumes are just too small.
cheers
- Glenn
You can already pay for faster routes with e.g. Google's premium networking or Cloudflare's Argo. If I understand correctly, this offering just means that instead of having to procure your network in advance and lock in a particular price for all traffic, you can dynamically route traffic based on the cost of doing so.
[0] Wikipedia's definition: "Network neutrality, or simply net neutrality, is the principle that Internet service providers (ISPs) must treat all Internet communications equally, and not discriminate or charge differently based on user, content, website, platform, application, type of equipment, or method of communication."
My understanding of net neutrality was that it is supposed to be a restriction on consumer residential/wireless ISPs' actions.
If you had an effective choice in ISPs, you could pick the one with properties you cared for, like lower latency connections; early internet gaming networks had prefered dial-up providers that shaved about 10ms off of cross-US ping times.
Riot Games has built its own network from major ISPs to their data centers to support League of Legends. This is not a terribly latency sensitive game, but it is worth it for them to do it because players are more likely to keep playing if they are mad at themselves and their team for losing as opposed to being mad at "the internet".
I assume Riot puts the service together by working with wholesalers and they probably have an "auction" on a timescale of months or years to determine routes.
I'm not sure what kind of latency issues other folks experience, but they are typically constrained by speed of light. People with big content networks like Google are going to directly peer with people with big eyeball networks like AT&T. Presuming that peering relationship is functional, the interconnects are not going to be over-subscribed and latency is going to be optimal between any two given points with no quality of service required, nor really any change for a third party to influence the quality of the interconnect.
This all presumes the network use case is gaming which has orders of magnitude less throughput required than streaming high quality video which is a different problem altogether.
There are still IXPs in the world (i.e. Equinix) where one can move a meaningful amount of traffic and you're perfectly correct in that you'll still need to pay for transit for the tier 1s.
Distance / the speed of light has nothing to do with it. If you're in the US connecting to a US game server, neither you nor the game developers have control over what path the internet will choose to send your traffic there.
Your packets could be blocked by other packets being sent first. You could experience packet loss, which will create a rubber banding effect in game. You could experience a variation in packet delivery time, which makes it hard for client-side prediction to work effectively.
These issues frustrate gamers to the point where they'll stop playing if it's bad enough. Lag is hated by games and developers, so this problem is super worth solving. Imagine what it'd be worth to Epic Games to reduce their player churn by just 0.1% monthly... probably millions! This is a big market.
If you read up on Network Next, they solve this problem by sending your traffic on private networks where you do have more control over delivery and consistency, rather than public networks where you don't.
I'm really excited to see Network Next improve gameplay, and it's 100% not a non-starter.
Per my example, let's assume you're an AT&T customer and you're hitting a the Stadia service or maybe any random gaming service hosted on AWS. AT&T peers with those guys directly. The most likely source of latency is going to be an oversubscribed edge router on AT&T's network just upstream of the client connection rather than a peering router. Regardless, how is some implementation of QoS going to address that? Is this company going to convince eyeball networks to implement QoS on their behalf? For millions of dollars? I seriously doubt it.
It's a non-starter in the sense that it is neither technically nor economically feasible to implement in a way that will solve the broad set of problems likely to be the cause of packet loss and latency.
And yet, our tech is up and working right now, providing real reductions in latency and packet loss for our customers.
Regardless, if you want to engage in some kind of rational discourse, I'm perfectly willing. My assertion is that your product, based on the scant information I've seen about it has a very limited scope in which it might improve packet loss and latency for gamers.
I hypothesize that most packet loss and latency issues for gamers arise from poor wireless networking. Those aside, the very serious gamers are going to be using wired connections. If they experience packet loss and latency, there are a few other possible culprits. Namely, the shit hardware most folks are forced to use by companies like Arris. Beyond that, there are last mile physical plant issues, and beyond that, edge routing devices which are typically woefully over-provisioned. There is no "private network arbitrage" service that's going to aid in those issues which I suspect comprise something north of 99% of all latency and packet loss related issues that gamers experience.
So, yeah, make some empty promises, extort some game developers and enjoy the same fate that many, many other similar companies before you enjoyed. See: all the BGP optimization companies that came and went in the 2000-2010 timeframe.
Care to explain to me how your product helps with latency when the client's network provider and the server's network provider have a direct peering relationship?
I'd also like to know how your product is going to improve packet loss and latency within the client network provider's network?
Tackle those two and I have a host of other questions. Welcome to the thread.