ZealousApathy

Yeah that's not surprising, Frankfurt is a central internet hub in Europe so internet tends to follow towards Frankfurt.

We did a lot of testing of different combinations of server tick rate and latency, to try and find a combination that minimized the perception of peeker's advantage (because you can never eliminate it, only minimize it). We settled on 128hz/35ms being an achievable target. We think it's "ok" up towards 50ms, but peeker's advantage becomes more and more prominent the higher your latency. That's why we generally aim for: 70% < 35ms, 90% < 50ms.

We'd like 70% of our players to be below 35ms, and 90% to be below 50ms. Achieving 35ms universally is impossible. So we'll always be happy to look for ways to improve those numbers, but need to balance improving those numbers against the cost (financial cost and impact to matchmaking). We're already greatly exceeding 70% in the European shard.

So there are always going to be ping discrepancies between players, but a hypothetical 15ms ping difference is only one aspect of what it takes to win any encounter; ability usage, knowing what angles to hold, game sense, team coordination, and honestly a bit of luck go into winning an encounter.

No plans for more coverage in North America right now, sorry!

A quokka: https://en.wikipedia.org/wiki/Quokka

I think we all know LATAM isn't where it could be right now, so we'll be looking closely at what we can do to improve LATAM in 2021.

No plans right now, we try to use AWS where we can and they're only in Mumbai. That's not to say we'd never think about another server location in India, but we'd need to see strong engagement from the Indian player base to be confident that a) matchmaking would still be viable and b) it was worth the investment (because we'd have to build a solution outside AWS, e.g. Microsoft Azure is in Chennai, but we don't support Azure right now).

Even if both Bahrain and Mumbai were in the same shard we'd already be splitting the player population because we prefer the server with lower latency. Yes, there might be some players like your friend in Dubai who have decent latency to both, but with a 10ms difference hopefully they always end up on Bahrain. The reason that Bahrain is on EU is that at launch, we didn't want to block Middle Eastern players from playing the game, and we thought that being on EU made the most sense while we waited for Bahrain to come online. Now that it has, we don't want Middle Eastern players to lose the ability to play with their friends in EU.

Ultimately, I believe we have the player population to make this layout work.

I don't have firm plans that I can talk about but it's on my priority list for next year. Real talk, the problem is that infrastructure is expensive in Latin America, without much coverage from public cloud providers (e.g. AWS is only in Sao Paulo, which is why most games are only hosted there in Latin America). We think we can do better, but I want to under-promise and over-deliver.

Yep, you've identified the primary case for why we'd put servers in Milan. It's on our radar, but we don't have any work planned yet.

The goal is to balance two tensions: more servers to lower latency, but fewer servers to concentrate the player base and improve matchmaking.

I felt it was better to "pull" players in from Florida to Atlanta, than to push more central players out to Miami. Purely a judgement call on my part, that a central location will better serve more players.

No plans right now. Your use case makes sense, but providing that support will only impact a very small part of the player base (e.g. those currently traveling outside their shard, or players with friends across the world who want to play together and have terrible latency).

I said this elsewhere:

The goal is to balance two tensions: more servers to lower latency, but fewer servers to concentrate the player base and improve matchmaking.

We don't have any plans to put servers in Milan right now, but we've got that option if we think that the player base fragmentation is worth the reduction in latency for some players.

Most of our servers are hosted in AWS. In some locations, generally the "2" locations as you've noted, we chose to expose some servers via AWS Edge in addition to Riot Direct, because AWS Edge has better routing for some players. Over time, we'll improve Riot Direct's routing and deprecate the AWS Edge servers.

Re: getting better latency to San Jose than Portland, the technical reason is usually ISPs peering in an unexpected way. Can you tell me what ISP you're on?

Re: expanding into Seattle, no plans, adding servers in Seattle would be a small incremental improvement to a relatively small player population that are mostly already being served well out of Portland. The goal is to balance two tensions: more servers to lower latency, but fewer servers to concentrate the player base and improve matchmaking.

I worked on the infra plan, AMA (within reason)

I'm going to DM you to get some details

Apologies, my Indian geography isn't great. Is Mangalore the city you live in, and BBNL the ISP?

Sorry buddy 😔 Unlikely that there'd ever be servers there, it's not densely populated enough

It's not unusual for particular ISPs in particular regions to have sub-optimal routing. We'll compile direct user feedback like this and also look at our player telemetry to see what changes need to be made. Unfortunately it also requires the ISPs to cooperate which isn't always fast!

I'm going to add Bangalore/India/Airtel to our list.