Crissic Migrating Jacksonvile - Much Downtime Planned

[drmike]

Email just went out

DATACENTER MIGRATION - SERVICE IMPACTING

Location: Jacksonville, FL (JAX)
Date: Saturday, August 22nd, 2015
Start Time: 5:00PM EDT
End Time: 3:00AM EDT

Description:
In order to offer better connectivity and to offer an improved level of hands-on support, your equipment is being moved to QuadraNet’s Miami facility. You should expect a total of one [B]outage, up to 10 hours in length[/B]. Each hypervisor will be shut down gracefully at the start of the migration.

In order to minimize the risk of extended downtime, and ensure that equipment is safely and securely transported to the new facility, QuadraNet staff members will be on-site to complete this migration. The new facility is located in Miami, FL and has substantial cross-connect capacity to all major Miami internet exchanges and meet-me-rooms. If you have any problems with your connection after the maintenance window, or if you have any questions regarding the maintenance at any point, please contact our support department at [email protected].

We appreciate your understanding and patience during this maintenance and welcome all feedback.

Thank You,

Crissic Management Team
[email protected]

 

[DomainBop]

Posted 11 August by DomainBop

Question:  anyone want to take bets on how long it will be before the Jacksonville customers are migrated (probably to Miami or Dallas)?

I guess I have my answer now. 

 

[drmike]

They lost me at the first sentence:

In order to offer better connectivityIs that some dig against the past two plus years at GoRack? If Miami DC for QN was such bandwidth hotness, I'd probably have noticed a brand with traction there.  I haven't.  Does one exist? Maybe.  Bigfoot might exist too.

This isn't about better connectivity.  

 

[OSTKCabal]

QuadraNet's Dallas and Miami locations have mediocre networks at best. Can get much better value and connectivity elsewhere... like at GoRack, who they're leaving. Nothing to do with better connectivity at all, and QN knows it just as well as everyone else.

 

[danielm]

It's entirely about cutting costs, but they can't say that publicly without bothering customers. QN has space in Miami, so financial logic says move Crissic to that space. The problem with following financial logic is that it doesn't usually translate to customer satisfaction. 

I think this was expected by many, I'm just a bit surprised how soon it is happening.

 

[OSTKCabal]

It's entirely about cutting costs, but they can't say that publicly without bothering customers. QN has space in Miami, so financial logic says move Crissic to that space. The problem with following financial logic is that it doesn't usually translate to customer satisfaction. 

I think this was expected by many, I'm just a bit surprised how soon it is happening.

I can't recall a recent hosting provider acquisition that has ended up better for the customer. Unfortunately, that's just the way of the world I guess. It doesn't have to be, but it is.

 

[drmike]

No test IP either.   I am sure this buying audience [the Crissic customers] wants such since they are a more informed customer base.

 

[OSTKCabal]

Oh no, people might find out that their ping is universally worse.

 

[DomainBop]

No test IP either.   I am sure this buying audience [the Crissic customers] wants such since they are a more informed customer base.

Does a cage qualify as a data center (address "Attn: Cage 9")? 

Miami test IPs: 173.44.32.10 and 2607:ff48:0:a:b:c:d:666

 

[drmike]

I just was looking at BGP for that IP.

Cogent… Gtt/tinet… Telefonica… Telia  as the BW mix.

Most routes / traffic appear to be going over Cogent.

 

[drmike]

Does a cage qualify as a data center (address "Attn: Cage 9")? 

Yes, I think we had a thread where we equated data center to equal the same non definition  as cloud.  Then the sun came up and evaporated all the bullshit and we didn't even get a rainbow.  Now I am left with Skittles and a bad memory of what rainbows look like.

And someone stole my pretty unicorn.

 

[KuJoe]

I've got a few VPSs in GoRACK (including a Crissic VPS) and one in QuadraNet Miami so I ran some pings to see the latency difference. For IPv4 QuadraNet has better latency although in some cases it's only by a few milliseconds, for IPv6 GoRACK has better latency although, again, some cases it's only by a few milliseconds.

 GoRACK IPv4QuadraNet Miami IPv4GoRACK IPv6QuadraNet Miami IPv6Phoenix, AZ, USA (01)79.534ms61.342ms60.731ms62.238msPhoenix, AZ, USA (02)72.018ms68.738ms54.626ms57.430msLos Angeles, CA, USA (01)67.974ms64.367ms65.142ms72.845msLos Angeles, CA, USA (02)60.606ms77.947ms73.881ms64.317msLos Angeles, CA, USA (03)73.035ms63.950ms74.307ms68.172msLos Angeles, CA, USA (04)67.937ms63.626ms67.773ms68.404msCentenial, CO, USA47.337ms51.262ms66.995ms60.388msDenver, CO, USA48.454ms57.979ms69.142ms60.263msJacksonville, FL, USA (01)0.052ms7.471ms0.144ms59.990msJacksonville, FL, USA (02)0.422ms14.258ms5.924ms21.033msMiami, FL, USA (01)60.490ms0.845msPing failed.Ping failed.Miami, FL, USA (02)7.530ms0.020ms59.961ms0.084msAtlanta, GA, USA33.921ms31.631ms6.927ms15.217msChicago, IL, USA40.340ms34.704ms43.638ms43.188msKansas City, MO, USA55.325ms44.885ms53.940ms44.106msLenoir, NC, USA15.599ms28.708ms19.994ms23.745msCharolette, NC, USA35.820ms18.731ms33.905ms53.855msPiscataway, NJ, USA31.211ms31.484ms33.527ms34.417msLas Vegas, NV, USA68.197ms75.537ms83.595ms79.646msPortland, OR, USA131.417ms106.819ms70.798ms79.516msDallas, TX, USA (01)208.269ms32.456ms54.390ms36.878msDallas, TX, USA (02)45.617ms44.458ms39.934ms67.941msSeattle, WA, USA (01)135.564ms96.316msPing failed.Ping failed.Seattle, WA, USA (02)85.914ms91.245ms132.953ms85.354msVarna, Bulgaria152.959ms173.161ms142.529ms490.792msBeauharnois, Quebec, Canada87.273ms39.794ms34.221ms118.546msKitchener, Ontario, Canada58.651ms45.996ms57.675ms59.861msRoubaix, France158.451ms113.496ms105.184ms106.682msDusseldorf, Germany106.556ms125.082ms110.340ms115.728msFalkenstein, Germany110.652ms123.805ms124.882ms122.094msTokyo, Japan191.143ms174.834ms188.049ms173.495msRotterdam, Netherlands113.046ms106.382ms104.907ms161.085msNottingham, UK123.130ms101.729ms100.852ms101.746msSingapore (1)248.415ms236.540msPing failed.Ping failed.Singapore (2)282.352ms321.672ms276.467ms255.142msStockholm, Sweden132.535ms127.995ms130.591ms142.972msSydney, Australia228.025ms227.806ms236.012ms235.201msTampa, FL, USA53.972ms27.421ms41.429ms6.802ms Out of 38 test servers here are the totals (numbers are how many of the tests that IP had lower latency):

GoRACK IPv4: 13
QuadraNet IPv4: 25
GoRACK IPv6: 20
QuadraNet IPv6: 15

I basically just put my test IPs into my mass ping site (chk2.net) and threw the results into a spreadsheet so nothing scientific really and the numbers don't mean much since you can see some of the tests are within 1-2 milliseconds of each other so hardly noticeable. Interpret the data as you wish, I just wanted to throw out some data for people to review.

NOTE: The GoRACK IP I used was my Crissic IP so when it migrates I can rerun the test and post the results here to compare if it got better or worse (although latency is just a small part of network testing, it's something we can easily test).

 

[OSTKCabal]

Not afraid to admit when I'm wrong. Appreciate the testing, KuJoe. I'm kind of curious about the speeds people would be seeing but the ping times are pretty telling. I'm wondering if GoRack is putting a lot of their outbound traffic through Abovenet rather than NodesDirect or TATA Communications, as that could certainly account for at least part of the problem.

**Edit - based on the information given below, this post is fairly well irrelevant. At the time of posting, I was not aware that Crissic used its own blend nor that the IP @KuJoe used belonged to said Crissic blend and not GoRACK.

 

[SkylarM]

Not afraid to admit when I'm wrong. Appreciate the testing, KuJoe. I'm kind of curious about the speeds people would be seeing but the ping times are pretty telling. I'm wondering if GoRack is putting a lot of their outbound traffic through Abovenet rather than NodesDirect or TATA Communications, as that could certainly account for at least part of the problem.

Crissic never used GoRack for bandwidth, we went with NodesDirect directly. A lot of our traffic runs through Atrato though. Also keep in mind that GoRack is backhauled to Atlanta, and the only "on site" Bandwidth is Atrato, which still does wonky routing anyways. 

We always had a weird network flow, where more inbound was over NodesDirect, with more outbound by Atrato (could be the other way around, don't remember)

 

[SC-Daniel]

Appreciate the testing. I'm kind of curious about the speeds people would be seeing but the ping times are pretty telling. I'm wondering if GoRack is putting a lot of their outbound traffic through Abovenet rather than NodesDirect or TATA Communications, as that could certainly account for at least part of the problem.

Just a reminder, Crissic does not use the GoRACK network, but rather uses their own mix of Atrato/Hibernia and us (Nodes Direct).

The Nodes Direct network uses PCCW, Atrato, Zayo, Tata, and private peering. We do very little to influence a particular carrier to be used, but majority of our traffic prefers PCCW.

I also represent GoRACK, who has a slightly different blend, but a vast majority of the traffic prefers Tata.

Best of luck to those using Crissic. If you are using another provider in the GoRACK facility and feel like there is a suboptimal route being used, please reach out to me and I can find a better path if one exists.

KuJoe, I would be interested if you could send me your spreadsheet and test IP(s) so I can run some tests.

Thanks,

Daniel

 

[KuJoe]

I considered doing speed tests but that's not something I could automate or perform in a reasonable amount of time so I might do traceroutes instead so we can see the routes being taken.

 

[OSTKCabal]

Not afraid to admit when I'm wrong. Appreciate the testing, KuJoe. I'm kind of curious about the speeds people would be seeing but the ping times are pretty telling. I'm wondering if GoRack is putting a lot of their outbound traffic through Abovenet rather than NodesDirect or TATA Communications, as that could certainly account for at least part of the problem.

Crissic never used GoRack for bandwidth, we went with NodesDirect directly. A lot of our traffic runs through Atrato though. Also keep in mind that GoRack is backhauled to Atlanta, and the only "on site" Bandwidth is Atrato, which still does wonky routing anyways. 

At the time I posted, I wasn't aware that KuJoe used the Crissic IP, not one directly with GoRACK. That's my mistake too.

 

[KuJoe]

Appreciate the testing. I'm kind of curious about the speeds people would be seeing but the ping times are pretty telling. I'm wondering if GoRack is putting a lot of their outbound traffic through Abovenet rather than NodesDirect or TATA Communications, as that could certainly account for at least part of the problem.

Just a reminder, Crissic does not use the GoRACK network, but rather uses their own mix of Atrato/Hibernia and us (Nodes Direct).

The Nodes Direct network uses PCCW, Atrato, Zayo, Tata, and private peering. We do very little to influence a particular carrier to be used, but majority of our traffic prefers PCCW.

I also represent GoRACK, who has a slightly different blend, but a vast majority of the traffic prefers Tata.

Best of luck to those using Crissic. If you are using another provider in the GoRACK facility and feel like there is a suboptimal route being used, please reach out to me and I can find a better path if one exists.

KuJoe, I would be interested if you could send me your spreadsheet and test IP(s) so I can run some tests.

Thanks,

Daniel

Sent.

I do have another VPS in GoRACK (NodeServ) so maybe I should run the same tests with that IP and add it to the spreadsheet. I also have another VPS in Miami not in QuadraNet so maybe I should also add that for comparison...

 

[DomainBop]

I've got a few VPSs in GoRACK (including a Crissic VPS) and one in QuadraNet Miami so I ran some pings to see the latency difference. For IPv4 QuadraNet has better latency although in some cases it's only by a few milliseconds, for IPv6 GoRACK has better latency although, again, some cases it's only by a few milliseconds.

I don't see any South America/ Central America / Caribbean in that list which is one of the prime reasons people choose that corner of the country, so I'll fill in the (IPv4) blanks:

Crissic
Argentina - Buenos Aires 192.175ms
Brazil - Sao Paulo 126.233ms
Brazil - Porto Alegre 199.143ms
Brazil - Rio de Janeiro 140.906ms
Chile - Vina del Mar 128.745ms
Colombia - Medellin 77,419ms
Costa Rica - San Jose 52.598ms
Panama - Panama City 41.674ms
Curacao -Willemstad 45.477 ms
Bahamas - Nassau 13.911ms
Puerto Rico - San Juan 38.545ms

Quadranet
Argentina - Buenos Aires 131.216ms
Brazil - Sao Paulo 111.5803ms
Brazil - Porto Alegre 148,248ms
Brazil - Rio de Janeiro 101.594ms
Chile - Vina del Mar 109.300ms
Colombia - Medellin 76.667ms
Costa Rica - San Jose 49.853ms
Panama - Panama City 34.824ms
Curacao -Willemstad 39.663 ms
Bahamas - Nassau 8.787ms
Puerto Rico - San Juan 29.818ms

 

[OSTKCabal]

DomainBop,

Are you able to provide a couple traceroutes? I'm curious what the major outbound routes/carriers are being used with Crissic versus Quadranet.