Fastpass+, a solution to "overwhelmingly negative" responses from families

[Gabe1]

I don't know anything about Internet-tech stuff. But I know Disney needs to do a better job because it is slow a lot and basically unusable sometimes, like today.

When Brody told Quint, "You're gonna need a bigger boat," he had far less knowledge about sharking...but he knew they needed a bigger boat.

Disney needs a bigger router (or whatever.)

Excellently stated. We don't need tech verbiage to be painfully aware Disney is swimming upstream with their WiFi. Worse they went from problems at Christmas to now and it is still sucidoodles.

 

[englanddg]

Well, the reason I brought up the edge / peering management is...

That's where the real cost of networking is.

So, it's a cost control decision. Which, I am pretty sure is not being made by their IT people.

Reporting and estimating network usage is fairly easy, even on a large network.

Adding bandwidth, at the levels they are going to need, not so much.

Here's an example, an OC-3 carries about 150 megs, with about 6 megs overhead, so ~144 megs usable. An OC-3 costs 25 - 30k PER MONTH.

Most home cable modems today are pushing 10 - 20 megs...for 100 a month (with a huge cable infrastructure supporting it) for around 100 a month or less.

 

[luv]

It's not their router. In fact, the traffic shaping on the routers and QoS management is part of the problem.

You know those dumb women who don't know ANYTHING about tech stuff and use Apple products because they're so dumb they can't figure out others? The ones who say their kids have to do tech stuff for them? That's me.

I hit the "like" on your post and will assume you're right. I don't understand any of it. I just know they need a bigger boat.

 

[flynnibus]

It's not their router. In fact, the traffic shaping on the routers and QoS management is part of the problem.

I've found the Wifi drops many times, not because of signal, but due to DHCP lease expiration. This means that they have QoS (Quality of Service) set so low that unless you are constantly using network signal, then you lose your connection and must reconnect.

They are doing this for a variety of reasons, but I suspect the primary one is a bottleneck at what networking geeks would call "the edge"...in other words, they are attempting to manage the amount of traffic passing between their various ISP peering relationship.

these two paragraphs and their series of connections make no sense.

DHCP has nothing to do with traffic shaping except for the notion of ensuring those packets get through (which is a local subnet function.. DHCP is broadcast) and potentially which group you are seen as for policy behaviors.

Your wifi connectivty disconnecting has nothing to with dhcp or upstream traffic shaping. Your end to end throughput? Yes.. wifi link? no

Do people forget their OSI model so quickly?

Even simple ipconfig will show you the DHCP lease time.. and your client will try to renew it NUMEROUS times before it actually expires. But if your physical link goes down.. your lease is tossed out by your client regardless of expiration time. (tho most will try to use the same IP in their next dhcp request... but the server doesn't have to respect it).

If the problem were simply oversubscribed upstream network.. they can solve that overnight if they haven't hit their physical limits.. and if they have, 45 days you can have new fiber in. Choke points on upstream network for something like this is probably the higher level functions like proxying, filtering, inspection, etc. Simple bandwidth is cheap and fast... application aware network handling is slower and much more expensive to scale.

 

[englanddg]

You know those dumb women who don't know ANYTHING about tech stuff and use Apple products because they're so dumb they can't figure out others? The ones who say their kids have to do tech stuff for them? That's me.

I hit the "like" on your post and will assume you're right. I don't understand any of it. I just know they need a bigger boat.

The bigger boat analogy (or bigger router as you put it first) is absolutely correct, and an excellent analogy. That's exactly what they need.

Or, if you think of it like road traffic, they may have great roads in their neighborhood, but the connector to the interstate always gets clogged. So, therefore they do things like many big cities have done for rush hour.

They introduce "QoS" rules (Quality of Service)...so, going back to road traffic...they add stoplights at interstate entrances that turn on during peak time to meter how quickly cars merge into traffic. They close lanes down with preference to other traffic (like, WDW park management traffic overhead, which is very likely all running on the same physical network "roads"), etc...etc...

All that, when what really needs to be done is a larger road / better intersection built.

 

[englanddg]

these two paragraphs and their series of connections make no sense.

DHCP has nothing to do with traffic shaping except for the notion of ensuring those packets get through (which is a local subnet function.. DHCP is broadcast) and potentially which group you are seen as for policy behaviors.

Your wifi connectivty disconnecting has nothing to with dhcp or upstream traffic shaping. Your end to end throughput? Yes.. wifi link? no

Do people forget their OSI model so quickly?

Even simple ipconfig will show you the DHCP lease time.. and your client will try to renew it NUMEROUS times before it actually expires. But if your physical link goes down.. your lease is tossed out by your client regardless of expiration time. (tho most will try to use the same IP in their next dhcp request... but the server doesn't have to respect it).

If the problem were simply oversubscribed upstream network.. they can solve that overnight if they haven't hit their physical limits.. and if they have, 45 days you can have new fiber in. Choke points on upstream network for something like this is probably the higher level functions like proxying, filtering, inspection, etc. Simple bandwidth is cheap and fast... application aware network handling is slower and much more expensive to scale.

No, what I observed was that the Wifi link AND DHCP lease did NOT drop...however, routing to the outside world did. When I released and renewed, my connection would start working again. If I kept active streams (for example, played a youtube video or something whilst working on another tab), I could keep stable, and relatively good network connection to the outside world. However, if I say, got up to go get a soda, and left the machine unattended for even a short period, they would stop routing my traffic to the cloud...though I still had a lease, and could ping my gateway, etc...

And yes, you are correct. I misspoke (but no, I haven't forgotten my OSI model <grin>) I wasn't losing the lease, but it's like if they saw no traffic coming from my IP for a period, they just shut it down as far as routing it on.

 

[flynnibus]

Well, the reason I brought up the edge / peering management is...

That's where the real cost of networking is.

So, it's a cost control decision. Which, I am pretty sure is not being made by their IT people.

Reporting and estimating network usage is fairly easy, even on a large network.

Adding bandwidth, at the levels they are going to need, not so much.

Here's an example, an OC-3 carries about 150 megs, with about 6 megs overhead, so ~144 megs usable. An OC-3 costs 25 - 30k PER MONTH.

Most home cable modems today are pushing 10 - 20 megs...for 100 a month (with a huge cable infrastructure supporting it) for around 100 a month or less.

OC-3 is so 2000... that was a big pipe back then. But classic single wave fiber as high capacity network is way out of date.

A site like Disney is going to have heavy MetroWAN and/or DWDM... a phat MPLS network with their nodes right in the data peering exchanges for traffic hand-off.

 

[PhotoDave219]

I don't know anything about Internet-tech stuff. But I know Disney needs to do a better job because it is slow a lot and basically unusable sometimes, like today.

When Brody told Quint, "You're gonna need a bigger boat," he had far less knowledge about sharking...but he knew they needed a bigger boat.

Disney needs a bigger router (or whatever.)

QFT. "Sometimes that shark he looks right into ya. Right into your eyes. And, you know, the thing about a shark... he's got lifeless eyes. Black eyes. Like a doll's eyes. When he comes at ya, doesn't seem to be living... until he bites ya, and those black eyes roll over white and then... ah then you hear that terrible high-pitched screamin'. The ocean turns red, and despite all the poundin' and the hollerin', they all come in and they... rip you to pieces."

 

[flynnibus]

No, what I observed was that the Wifi link (DHCP lease) did NOT drop...however, routing to the outside world did. When I released and renewed, my connection would start working again. If I kept active streams (for example, played a youtube video or something whilst working on another tab), I could keep stable, and relatively good network connection to the outside world. However, if I say, got up to go get a soda, and left the machine unattended for even a short period, they would stop routing my traffic to the cloud...though I still had a lease, and could ping my gateway, etc...

That's a function of shaping, filtering, and potentially proxying. Yes, you do such things for managing your outflow traffic... for security, auditing, and bandwidth management. But you don't set it up so things 'die'.. symptoms like that show the system is overtaxed. If the problem were simply bandwidth demand.. they could fix that in a snap.

Bandwidth is an ondemand service for most well equipped sites these days.

 

[englanddg]

That's a function of shaping, filtering, and potentially proxying. Yes, you do such things for managing your outflow traffic... for security, auditing, and bandwidth management. But you don't set it up so things 'die'.. symptoms like that show the system is overtaxed. If the problem were simply bandwidth demand.. they could fix that in a snap.

Bandwidth is an ondemand service for most well equipped sites these days.

Hence why I brought up OC3s...I wonder exactly what they have in terms of peer ISPs...? As you said, it's a simple solution, until you are the IT guy going to management and saying "Oh...yeah, well...we need 40 of these 30k per month lines."

I'm not sure how well built out commercial ISPs are on their property...so, they very well could have a topology that sticks most traffic into a few scattered datacenters onsite and to the outside world. Their internal bandwidth could be more than enough to handle the demand, but not enough to handle all the external requests.

 

[Gabe1]

Well, the reason I brought up the edge / peering management is...

That's where the real cost of networking is.

So, it's a cost control decision. Which, I am pretty sure is not being made by their IT people.

Reporting and estimating network usage is fairly easy, even on a large network.

Adding bandwidth, at the levels they are going to need, not so much.

Here's an example, an OC-3 carries about 150 megs, with about 6 megs overhead, so ~144 megs usable. An OC-3 costs 25 - 30k PER MONTH.

Most home cable modems today are pushing 10 - 20 megs...for 100 a month (with a huge cable infrastructure supporting it) for around 100 a month or less.

Ahhhhhhhhhhh!

Like before Splash went down for rehab, didn't need an engineer to spout blah blah blah for why the interior wasn't working or why chunks of the mountain were crashing down. It was broke, reliability awful, needed attention and out.

WiFi reliability is sucidoodles, they need to attend to it. They haven't. Need to get off behinds and fix it.

 

[englanddg]

OC-3 is so 2000... that was a big pipe back then. But classic single wave fiber as high capacity network is way out of date.

A site like Disney is going to have heavy MetroWAN and/or DWDM... a phat MPLS network with their nodes right in the data peering exchanges for traffic hand-off.

I'd be curious to know what they have. The last major network I worked on was in 2001, so...yes, it is 2000. I'm mostly a developer now. I colo in a SOC 1 datacenter so I don't have to learn these things in great detail.

I still picture them with a bottleneck to the outside world...whilst their internet WAN is fine.

 

[PhotoDave219]

They need a bigger pipe. They have great signal everywhere; theyve fixed that problem. My wifi doesnt drop - it just doesnt move.

Their bandwidth cannot handle the current guest load. What's going to happen when we role this all out?

 

[flynnibus]

Hence why I brought up OC3s...I wonder exactly what they have in terms of peer ISPs...? As you said, it's a simple solution, until you are the IT guy going to management and saying "Oh...yeah, well...we need 40 of these 30k per month lines."

I'm not sure how well built out commercial ISPs are on their property...so, they very well could have a topology that sticks most traffic into a few scattered datacenters onsite and to the outside world. Their internal bandwidth could be more than enough to handle the demand, but not enough to handle all the external requests.

Your mindset is dated. What you are talking about is the days of paying for telecom bandwidth from an ISP. Now.. you buy pipes.. and you lay those pipes to exchanges.. and peer in the datacenter. What was once the land purely of ISPs.. peering in the datacenter... is how all major subscribers work now. So I don't buy an OC-3 from AT&T, and separate one from BBN, etc.. You get into the datacenter, and you make arrangements for direct interfaces locally. A big organization gets fiber back to the DC.. or simply buys MLPS connectivity.. over which they logically lay out their pipes.. and then hang a sweet high speed router at the end of some of those pipes in the datacenter.. and then peer locally in the DC.

Your 'edge' is in a datacenter where you have extreme access to peering and bandwidth. Then you long haul 'inside' connectivity back to your sites over MLPS links over whatever capacity physical link you need. You manage bandwidth of your WAN through your MLPS pipes between the sites. When you buy that MLPS connectivity from the big boys.. you can get insane DWDM fiber.. and they just fill out an order for whatever logical circuits and capacities you want. Takes longer to provision through the bureaucracy than it does to actually enable the capacity.

Bandwidth and connectivity are cheap relative these days.. it's the awareness stuff that doesn't scale as easily as the pipes themselves.

 

[flynnibus]

They need a bigger pipe. They have great signal everywhere; theyve fixed that problem. My wifi doesnt drop - it just doesnt move.

It's not that simple.. even tho you have radio signal - that doesn't mean you are successfully getting through.

WiFi channels are pretty low capacity.. if you have too many people on a channel.. the efficiency of getting through clearly is hurt.

It's like being on a conference call.. and everyone trying to talk at once. All of you see you have a solid phone connection - but none of you can get your message through.

So you can have great signal - but if you can't have a reliable conversation because there are too many people trying to fight for the same limited time.. you are slow and wait.

This is where the real complexity of high capacity WiFi networks comes in. It's not about radio coverage.. it's about enough radios to handle the # of clients over the area you need.

You have to solve those LAN problems.. then you move into WAN problems where you start having so much traffic consolidated and being able to handle those loads, etc.

 

[PhotoDave219]

It's not that simple.. even tho you have radio signal - that doesn't mean you are successfully getting through.

WiFi channels are pretty low capacity.. if you have too many people on a channel.. the efficiency of getting through clearly is hurt.

It's like being on a conference call.. and everyone trying to talk at once. All of you see you have a solid phone connection - but none of you can get your message through.

So you can have great signal - but if you can't have a reliable conversation because there are too many people trying to fight for the same limited time.. you are slow and wait.

This is where the real complexity of high capacity WiFi networks comes in. It's not about radio coverage.. it's about enough radios to handle the # of clients over the area you need.

You have to solve those LAN problems.. then you move into WAN problems where you start having so much traffic consolidated and being able to handle those loads, etc.

My overwhelming point is that Disney already has more traffic then it can handle. And we havent gotten started.

 

[flynnibus]

My overwhelming point is that Disney already has more traffic then it can handle. And we havent gotten started.

I don't disagree - but where/how they are failing and how to fix it should probably be left to the pros

 

[PhotoDave219]

I don't disagree - but where/how they are failing and how to fix it should probably be left to the pros

Im not saying how they should fix it, rather I'm pointing out the fatal flaw in their $2 billion+ endevour that needs fixing lest they lose that bet

 

[englanddg]

Your mindset is dated. What you are talking about is the days of paying for telecom bandwidth from an ISP. Now.. you buy pipes.. and you lay those pipes to exchanges.. and peer in the datacenter. What was once the land purely of ISPs.. peering in the datacenter... is how all major subscribers work now. So I don't buy an OC-3 from AT&T, and separate one from BBN, etc.. You get into the datacenter, and you make arrangements for direct interfaces locally. A big organization gets fiber back to the DC.. or simply buys MLPS connectivity.. over which they logically lay out their pipes.. and then hang a sweet high speed router at the end of some of those pipes in the datacenter.. and then peer locally in the DC.

Your 'edge' is in a datacenter where you have extreme access to peering and bandwidth. Then you long haul 'inside' connectivity back to your sites over MLPS links over whatever capacity physical link you need. You manage bandwidth of your WAN through your MLPS pipes between the sites. When you buy that MLPS connectivity from the big boys.. you can get insane DWDM fiber.. and they just fill out an order for whatever logical circuits and capacities you want. Takes longer to provision through the bureaucracy than it does to actually enable the capacity.

Bandwidth and connectivity are cheap relative these days.. it's the awareness stuff that doesn't scale as easily as the pipes themselves.

Fascinating how far it's come in a decade. I haven't touched networking in any detail for years. My current largest network is extremely simple, with the most complex aspect being OpenVPN.

In the early 2000's I was working for DoD, and we still had sites limited to ISDN in many cases...

But what you are saying makes sense, it's the "burst" capability of the DC...

But, there still needs to be peering connections, or am I wrong? And those peering connections cost money for throughput?

In addition, logical can't trump physical. Could it be that they have physically maxxed out the available bandwidth (I'd imagine that difficult to do, even for someone planning a basic DC, but possible).

I was simply flummoxed why I could see the connection where they handed it off simply drop, unless I maintained a constant stream in the background. It rendered devices like iPads/Droid phones (at the time) nearly useless without "reconnecting" to the wifi, but my laptop, I could easily create network activity in another tab on the browser to keep my session active.

 

[englanddg]

A touch off topic, but I will say the NICE systems they use (caps mean the company, not emphasis) are quite impressive, and I am very impressed with how Disney manages their phone systems currently.