You have made enough money to move the NA/EU servers

[ReplicaX]

Man as I stated before these problems can easily be fixed with a VPN, 4 bucks for pingzapper, people can skip out on a happy meal or 2 once a month.

I don't need to know this when I have used chained VPNs since the late 90s and early 2000s for Japanese Arcade Machine Networks dealing with Sega.

You are also talking to a Player that is less than 300 Miles (483Km) to the DC. 200 Miles (386Km) of my Route being inside my fiber network of 15ms avg (My ISP is actually a T1 Provider). Then hitting Level 3 and SE's Ormuco the rest at 100ms with spikes upto 500ms at Destination. I chain through Quebec City to hit Montreal from the North East side to bypass Level 3 getting an avg of 25ms and 0 spikes. Likely the route you use coming from the UK by default.

Level 3 owns all the main backbones in Montreal coming in from New York and Chicago's main routing. That is basically most of the US East Coast and Middle States of this Country unless you have a odd ball route. Cogent covering a good portion of the West.

Yet you are in the UK over 3,000 Miles (4828Km) away and think your ISP is why you have a great connection. This is your pure luck being distance + # of literal hops with your ISP not controlling your routing outside its network.

I have 7 hops to the DC, 4 of which are inside Montreal without chaining VPNs. You likely have over 20.

[Ephier]

I don't need to know this when I have used chained VPNs since the late 90s and early 2000s for Japanese Arcade Machine Networks dealing with Sega.

You are also talking to a Player that is less than 300 Miles (483Km) to the DC. 200 Miles (386Km) of my Route being inside my fiber network of 15ms avg (My ISP is actually a T1 Provider). Then hitting Level 3 and SE's Ormuco the rest at 100ms with spikes upto 500ms at Destination. I chain through Quebec City to hit Montreal from the North East side to bypass Level 3 getting an avg of 25ms and 0 spikes. Likely the route you use coming from the UK by default.

Level 3 owns all the main backbones in Montreal coming in from New York and Chicago's main routing. That is basically most of the US East Coast and Middle States of this Country unless you have a odd ball route. Cogent covering a good portion of the West.

Yet you are in the UK over 3,000 Miles (4828Km) away and think your ISP is why you have a great connection. This is your pure luck being distance + # of literal hops with your ISP not controlling your routing outside its network.

I have 7 hops to the DC, 4 of which are inside Montreal without chaining VPNs. You likely have over 20.

I did not know that Germany was in the UK.

And I have 10 hops thank you.

[Raist]

Required reading for anyone who wants to have an actual intelligent debate about this topic, rather than speaking out of your own butthole: 10 Places You Don't Want a Data Center

From this article, it's very clear that California (huge earthquake hotzone, stilling waiting on a long over "big one") and New York City (the densest metropolitan area in the country) are terrible locations for a major data center. Montreal is actually a surprisingly good choice based on its location (it's deceptively close to New York City, Boston, and Chicago), relatively infrequent natural disasters, and an unusually "high-tech" city given its size and population. It also helps that it's roughly equidistant to both California and Europe. Given that they're adding a new EU dedicated data center in Europe, I would highly recommend either Seattle or Vancouver for a "west coast" data center.

Love the tag on this thread about the Dunning-Kruger effect, though it really should apply to 90% of threads in this forum.

Except that Vancouver has even more restricted routing options than Montreal.

Level3

Tata

Cogent


SE likely initially looked at Montreal because of the EIDOS aquisition, and then found out on paper it was actually a pretty wise choice for what they were looking to do. Europe has undersea access that drops practically right outside their doorstep. Otherwise, they might have been able to tap into their footprints elsewhere in the states, but the ones that likely had anything remotely like the inrastructure in place that they needed would have been Southern Cali or New York---their Texas footprint was more along the lines of a warehouse for product fulfillment than something you would convert to a massive server farm.

Something that seems to be getting missed here also. This isn't just a room of simple servers. It is racks and racks of servers and various storage arrays and network devices. There is a tiered topology of frontend/backend flow that must be maintained, monitored, and continually tweaked. Years ago I found an article that had an overview of their internal messaging system and it was massive. Wish I could find it again just to have the visual for you guys. These things are not the kind of thing you just load on a U-Haul and carry it across the country.

Edit:
Crud.. found the thread in the XI forums with the link to SE's message system layout, but the link is dead--the only PDF still up is financial data. I did however find a pic in that same thread for another data center that might help make the point (granted, SE is likely using blade servers... but still, the point is it is a lot of hardware involved):


And here's a diagram from Cisco:

(Blade Centers in the Data Center--Overview)

[Raist]

So you continue to not understand that SE has their own Provider(s) that is directly connected to Level 3. You still fail to understand that SE also has that responsibility through the Provider they use.

Level3 is not the ONLY peering/transit provider available. Ormuco (SE's ISP) actually has a very robust peering agreement with not only the PSInet affiliate of Cogent, but also TATA.

(http://bgp.he.net/AS20324)

Here's the breakdown of those ASN's:
AS6453 TATA COMMUNICATIONS (AMERICA) INC
AS3356 Level 3 Communications, Inc.
AS3257 Tinet SpA
AS174 Cogent Communications

Perhaps if you were following these issues more closely, you would know more about these other peers available for them. Level3, TATA, and Cogent are continually brought up over and over in nearly all these ongoing threads. And it has been shown time and time again that there are big issues at the exchange points to these third-party networks. Here's the path I take, and atm there are no Level3 hops, and everytime TWC switches me to them I quickly get taken off of them:

  0  G75VXLAP [10.10.100.10]
  1  LPTSRV [10.10.100.1]
  2  cpe-075-176-160-001.sc.res.rr.com [75.176.160.1]
  3  cpe-024-031-198-009.sc.res.rr.com [24.31.198.9]
  4  24.31.196.212
  5  be33.chrcnctr01r.southeast.rr.com [24.93.64.182]
  6  bu-ether14.atlngamq46w-bcr00.tbone.rr.com [66.109.6.82]
  7  107.14.19.99
  8  te0-0-0-10.ccr21.atl02.atlas.cogentco.com [154.54.12.109]
  9  be2050.ccr41.atl01.atlas.cogentco.com [154.54.0.165]
 10  be2168.ccr21.dca01.atlas.cogentco.com [154.54.31.94]
 11  be2148.ccr41.jfk02.atlas.cogentco.com [154.54.31.118]
 12  be2106.ccr21.alb02.atlas.cogentco.com [154.54.3.50]
 13  be2088.ccr21.ymq02.atlas.cogentco.com [154.54.43.17]
 14  38.122.42.34
 15  192.34.76.10
 16  199.91.189.242
 17  199.91.189.30

And no... so long as TWC is staying on top of my route like they have been since I started working with them months ago on the issues, I don't have all the lag and disconnects everyone else experiences. And I'm coming from South Carolina.

[Id_Slayer]

Kansas might be a nice location if not chicago, for NA servers since Google has their gigabit net service available there, clearly there is enough internet backbone infrastructure there for servers. I imagine Saint Louis is the mid point between kansas city and chicago as well, so that may be an option worth looking into, too. :O

[Illya]

Required reading for anyone who wants to have an actual intelligent debate about this topic, rather than speaking out of your own butthole: 10 Places You Don't Want a Data Center

From this article, it's very clear that California (huge earthquake hotzone, stilling waiting on a long over "big one") and New York City (the densest metropolitan area in the country) are terrible locations for a major data center. Montreal is actually a surprisingly good choice based on its location (it's deceptively close to New York City, Boston, and Chicago), relatively infrequent natural disasters, and an unusually "high-tech" city given its size and population. It also helps that it's roughly equidistant to both California and Europe. Given that they're adding a new EU dedicated data center in Europe, I would highly recommend either Seattle or Vancouver for a "west coast" data center.

Love the tag on this thread about the Dunning-Kruger effect, though it really should apply to 90% of threads in this forum.

Wow, you sure are smart. I wish I could be as smart as you. I bet you're an IT manager who's in charge of datacenter operations for a large Internet company, or at least have extensive experience in MMO datacenter operations. You're so smart that you know about the Dunning-Kruger effect, you're like a genius or something. I mean, your informative link suggests a small town in Virginia as the ideal datacenter location, so I bet lots of big MMOs locate their datacenters in super-remote areas like that.

For example, a little company called Blizzard you might have heard of operates an obscure MMO called World of Warcraft. Surely they would follow the precepts laid down by the blatant marketing document for a hosting company that you linked as the Gospel of Datacenters. Hmmm...it turns out their North American datacenters are located in Los Angeles, CA and Phoenix, AZ (West) and Chicago, IL and New York City, NY (East).

Well then. I think the person speaking out of their butthole is actually you. Are you Tom Deaderick or something? Because if you're trying to sell Squenix on your amazing Tier III datacenter operations, I suggest you contact them directly rather than post on this forum.

[ReplicaX]

Perhaps if you were following these issues more closely.

I'm aware of this and my Provider does this for me now. I had the level 3 issues in Feb like many others. Here is my current hops outside of my network for proof:

xe-9-0-0.mtl10.ip4.gtt.net [141.136.105.162]
ormuco-gw.ip4.gtt.net [216.221.156.110]
[192.34.76.2]
[199.91.189.234]
[199.91.189.38] Avg: 24ms Min: 23ms  Max: 25ms Cur: 24

I'm not familiar with Global Telecom & Technology however they have been good since my provider rerouted me through them. 20 to 40ms at destination with little to fluxuations besides the recent DDoS attacks to the servers.

In February all my issues were from the level 3 to Ormuco peer exchange. Why I used a VPN in Quebec City until my ISP changed my routing.

My question to you is simple though. Regardless of Ormuco's agreements do you not feel Ormuco has responsibility with that peer exchange between them and Level 3 and working with them to resolve it?

My point was rather simple. It's not one Company or ISP to blame, its those connected to them as well. My Provider and SE's Provider are responsible to resolve these issues with Level 3. After months my Provider got no where with them and used alternate routing in Montreal as a solution via another Company.

Edit:

AS3257 Tinet SpA

Seems they were bought out by Global Telecom & Technology last year.

[Raist]

I'm aware of this and my Provider does this for me now. I had the level 3 issues in Feb like many others. Here is my current hops outside of my network for proof:

xe-9-0-0.mtl10.ip4.gtt.net [141.136.105.162]
ormuco-gw.ip4.gtt.net [216.221.156.110]
[192.34.76.2]
[199.91.189.234]
[199.91.189.38] Avg: 24ms Min: 23ms  Max: 25ms Cur: 24

I'm not familiar with Global Telecom & Technology however they have been good since my provider rerouted me through them. 20 to 40ms at destination with little to fluxuations besides the recent DDoS attacks to the servers.

In February all my issues were from the level 3 to Ormuco peer exchange. Why I used a VPN in Quebec City until my ISP changed my routing.

My question to you is simple though. Regardless of Ormuco's agreements do you not feel Ormuco has responsibility with that peer exchange between them and Level 3 and working with them to resolve it?

My point was rather simple. It's not one Company or ISP to blame, its those connected to them as well. My Provider and SE's Provider are responsible to resolve these issues with Level 3. After months my Provider got no where with them and used alternate routing in Montreal as a solution via another Company.

You were making the case that it is only level3 in play, which clearly is not the case and needed to be corrected.

As for Ormuco working with Level3, if it is found to be an issue with their peering agreement, then yes... they may need to revisit that agreement. However, in the majority of the cases I've seen in play there are actually tell-tale signs of congestion causing problems before it gets to the Ormuco handoff. The one time we spotted anything on their end (we being a TWC T3 tech and myself while looking at my route), it looked like an issue with their VLAN that was spiking +200 over the norm and we messaged their admin about it. It was fixed a couple days later. Otherwise, both when looking at mine and other people's routes we've been more consistently seeing too much jitter prior to exiting Level3, Cogent, or TATA on the Ormuco side.

GTT was bought up along with another groups lines if I remember right. Seem to recall seeing that a while back. Someone originally based out of Europe I think...some weird passing around where it was absorbed into another company, then that company was sold to them or some odd thing. The lines are currently owned by Tinet (German company? Their WhoIs is a mess), who peers with Ormuco directly.

ATM we are wrestling with our more local nodes here in the Carolinas. They were running me on Hilton Head's gateway for a while, then they switched me to Conway. It's a bit better then the mess that was starting at Hilton Head, but it's still not good... about 1/20 times we see a spike break 100ms within my first few hops while still in TWC territory. When it gets more frequent, I start to see the jitter on screen--but it's still not enough for rubberbanding or any timeouts.... yet. Just wish they would finally get Florence's head-end straightened out so I don't have to ping-pong all over the place like we're doing right now.

[Evangela]

I think this issue will be solved automatically in 2-3 years when the world's internet speed improve generally, like 1TB/s.

[MilesSaintboroguh]

I think this issue will be solved automatically in 2-3 years when the world's internet speed improve generally, like 1TB/s.

You mean everyone but the United States cause god forbid that the ISPs stop overcharging customers and only offer fast internet to people outside of cities.