On 1/25/22 6:11 AM, Josh Luthman wrote:

Mark,

Use the 12 foot ladder to get over the 10 foot paywall:

https://12ft.io/proxy?q=https%3A%2F%2Fwww.economist.com%2Fbusiness%2Fwill-th...

Yeah, sorry I didn't know what their paywall looked like since I subscribe Mike

On Tue, Jan 25, 2022 at 4:12 AM Mark Tinka <mark@tinka.africa> wrote:

On 1/19/22 21:52, Michael Thomas wrote:

> > There was an article in the Economist (sorry if it's paywalled) about > Dish entering the mobile market using an AWS backend. I don't think > that AWS brings much more than compute for the most part so I don't > really get why this would be a huge win. A win maybe, but a huge win? > I can certainly see that not having tons of legacy and accreted > inertia is big win, but that's true of any disruptor. In the end they > still need base stations, spectrum, backhaul and all of that to run > their network, right? > > Am I missing something, or is this mainly hype? > > Mike > > https://www.economist.com/business/will-the-cloud-business-eat-the-5g-teleco...

>

It's behind a pay wall, so can't read the entire article.

But AFAICT, the way AWS's 5G service works is that they can provide an entire solution (towers, backhaul, back-end, even SIM cards), or just the back-end.

I believe the latter is called Wavelength:

https://aws.amazon.com/wavelength/features/

I'd say it is a legitimate threat to traditional MNO's. One does not require to build a national-scale mobile network from scratch... if you can service a small community of some 500 people in a manner that is affordable to them, and gives you a nice return so you can buy some food at the end of each month, no reason why that is not a successful and sustainable model.

Heck, you probably don't even need to offer classic voice and SMS services. There are plenty of IP-based apps that will do this for you, and I know many people who have totally given up packages that include voice minutes and SMS messages, in favour of data-only services from their MNO. They are thriving.

So if a small mobile operator using AWS 5G as a back-end does not need to negotiate massive interconnect contracts and deals with other telco's in the area, and their handful of customers are fine with just Internet access as the only service, makes a lot of sense to me.

As I've been saying for a long time now, the telco model is a dying one. Customers only care about the problems we can help them solve, not whether we are a Tier-1, or how many TV shows were "Brought to you by..."

Mark.

On Tue, Jan 25, 2022 at 6:06 AM Mark Tinka <mark@tinka.africa> wrote:

On 1/25/22 15:45, Masataka Ohta wrote:

...

As is stated in free part of the article that:

The country’s three biggest carriers, AT&T, Verizon and T-Mobile, have offered 5G connectivity but in practice this differed little from the earlier 4G.

5G is nothing. That's all.

Considering the relatively decent performance of 4G/LTE, especially as fibre + wi-fi is more rife, particularly in the dense metropolitan areas that would be fibre-rich, with folk offloading a lot more of their traffic to wi-fi in lieu of GSM, I am still struggling to see the 5G use-case, outside of implementing 3GPP specs. for their own sake.

By one operator offering 5G, all other operators are forced to offer 5G. So they all end up spending billions to remain in the same place.

Ah well...

As far as AWS 5G goes, they don't have to be locked into just 5G. De-risking the back-end for non-traditional greenfields would allow them to even support GPRS if it made sense. It's all software and CPU.

I would say its all actually billions of $$ in spectrum and patent fees… hardware parts are a rounding error. The more interesting story is at&t 1. Tell everyone you are build an open source core network on openstack (lol) 2. Build it, put it into prod, then disown it into the linux foundation 3. Admit you built an albatross , then pay Azure to take it off your hands, and thus losing any control of your network.

Mark.

Mark Tinka wrote:

By one operator offering 5G, all other operators are forced to offer 5G. So they all end up spending billions to remain in the same place.

As 802 is for LANMAN, all we need is 802.11 for MAN maybe combined with IP mobility. Masataka Ohta

On Tue, Jan 25, 2022 at 10:11 AM Michael Thomas <mike@mtcc.com> wrote:

[...]

Since everybody has their own wifi it seems that federating all of them for pretty good coverage by a provider and charging a nominal fee to manage it would suit a lot of people needs. It doesn't need expensive spectrum and the real estate is "free". Basically a federation of "guestnets".

Mike

Which is pretty much what Xfinity is already offering to their subscribers; use your xfinity login to get onto the wifi access points of other xfinity users all around the country, relatively seamlessly. I'm sure other networks that provide their own CPE are likely to follow suit as well. Matt

On 1/25/22 11:44 AM, Matthew Petach wrote:

On Tue, Jan 25, 2022 at 10:11 AM Michael Thomas <mike@mtcc.com> wrote:

[...]

Since everybody has their own wifi it seems that federating all of them for pretty good coverage by a provider and charging a nominal fee to manage it would suit a lot of people needs. It doesn't need expensive spectrum and the real estate is "free". Basically a federation of "guestnets".

Mike

Which is pretty much what Xfinity is already offering to their subscribers; use your xfinity login to get onto the wifi access points of other xfinity users all around the country, relatively seamlessly.

I'm sure other networks that provide their own CPE are likely to follow suit as well.

What I was thinking of is more of "over the top" where I don't need to be an Xfinity customer (lot least of which is that I can't). It does seems that providing the CPE is inevitable though, but for a lot of people that's feature not a bug. Mike

On 1/25/22 11:44, Matthew Petach wrote:

Which is pretty much what Xfinity is already offering to their subscribers; use your xfinity login to get onto the wifi access points of other xfinity users all around the country, relatively seamlessly.

Which is a major hassle if you have your own APs and just want^H^H^H^H have no practical choice other than to use their Internet, as their CPE includes a radio that does nothing for you and that you can't turn off. I had to escalate through many layers of "support" to get to someone who could disable their enabled-by-default interference and noise generator after threatening to open up the box and cut the appropriate trace. -- Jay Hennigan - jay@west.net Network Engineering - CCIE #7880 503 897-8550 - WB6RDV

If you stay away from getting stuck in the word "cloud", there is lots of value for folk that choose to de-risk their infrastructure,

s/de-risk/re-risk/ it's just a different risk randy

For some folk, the risk of money cost outweighs the risk of loss of direct operational control.

Those folks also tend to learn hard lessons about what happens when the Magic Cloud provider fails in a way that isn't possible to anticipate because it's all black box. Saving 12 months of opex $ sounds great, except when you lose 18 months of opex $ in 2 days completely outside of your ability to control. On Wed, Jan 26, 2022 at 9:46 AM Mark Tinka <mark@tinka.africa> wrote:

On 1/26/22 16:41, Randy Bush wrote:

...

s/de-risk/re-risk/

it's just a different risk

I should have finished that sentence with "de-risk their infrastructure spend", because the actual risk is in having to spend money upfront to build the network.

For some folk, the risk of money cost outweighs the risk of loss of direct operational control.

Mark <= who run way from a suggestion to run iBGP route reflectors in AWS :-).

Mark.

On 1/26/22 7:10 AM, Tom Beecher wrote:

For some folk, the risk of money cost outweighs the risk of loss of direct operational control.

Those folks also tend to learn hard lessons about what happens when the Magic Cloud provider fails in a way that isn't possible to anticipate because it's all black box.

Saving 12 months of opex $ sounds great, except when you lose 18 months of opex $ in 2 days completely outside of your ability to control.

I think for the vast majority of cloud users they'd do a way worse job at uptime than the providers. Whether that applies to some telcos, I'm not sure. Mike

On 1/26/22 23:04, Christopher Morrow wrote:

It seems like some of the situation is:   "5g/mobile builds include a bunch more 'general machine' resources which offload a bunch of the work from what was dedicated appliances/etc."

Followed quickly by:   "Well, we don't have the resources/etc to design/build/run/maintain that sort of thing in the field"

In a bunch of mobile deployments (in the US at least) a lot of the work was done by some vendor already, so swapping one vendor for another isn't particularly new.    "Out with Nortel, in with Ericcsson!"

As to 'is this cloud?' or not, that's probably not super important? If the telco (as an example) could come to an agreement with ~bunches of local sysadmin shops who'd all cooperate and build/deploy 'the same thing' (from the goes-into and goes-outof perspective) a price points which would be palatable. I imagine the telcos would have taken that direction.  Instead, they choose to minimize the number of contracts and options and get cookie-cutter deployments.

Folk may grate at 'aws' or 'azure' or 'gcp' ... but really the telco folk (the customer in this case) is choosing someone to run infra for them, under contract with what they hope are appropriate SLO/SLA and repair properties. It certainly behooves them to think about failure scenarios, but that's what SLO/SLA are for, right? :) and offloading the methods of repair/avoidance is part of the contract process.

The classic MNO's will continue the same operational model, of contracting a vendor and giving them a floor in their building so they can operate the network. For them, it's a case of "if it works, don't fix it". What content are doing by getting in on the game is, really, to open up the industry to other players that may want to either run their own private cellular networks, or start new mobile businesses at whatever scale they can muster, without having to lay out a ton of fresh capital in terms of money and people, to get going. There is still room for both classic MNO and content-backed infrastructure. Which one wins out in the end is a matter of time, but it's great for the user, because they just want to get connected. They don't really care, anymore, who gives them that connection, as long as it works and doesn't break the bank. Mark.

On 1/26/22 17:10, Tom Beecher wrote:

Those folks also tend to learn hard lessons about what happens when the Magic Cloud provider fails in a way that isn't possible to anticipate because it's all black box.

Saving 12 months of opex $ sounds great, except when you lose 18 months of opex $ in 2 days completely outside of your ability to control.

I don't disagree. What this does, though, is democratize access into the industry. For a simple business model that is serving a small community with a handful of eyeballs, not trying to grow forever but put food on the table, it's somewhere to start. Mark.

On 1/27/22 09:32, Saku Ytti wrote:

I do disagree, if I understood the argument right. If the argument is 'cloud makes no business sense to anyone'.

I don't agree that cloud does not make business sense to anyone. There is a reason why Amazon, Microsoft and Google are milking it right now, so that is not even a discussion. What I do agree with is that the loss of control of operating your network yourself does present a risk. But that is a personal position, and has no bearing on the ultimate sensibility of offloading your infrastructure to a cloud that is likely to run it better than you, most of the year. It's one of the reasons I have no desire to work for an MNO as a hardcore engineer - I can't stomach the idea of being a vendor's project manager :-).

Doing the 1st server properly costs several million euros a year, since you need competent 24/7 staffing, with sick leaves, holidays (in 1st world countries where this is a thing) and attrition taken into account. Staff who can do infra, compute, storage, networking (that's 4 separate teams usually, each needing overhead for 24/7) who are comfortable with working nights.

This argument 'no one should be using x, x is a fad' happens when every new technology appears, literally people object to using paper and pen, as it's too convenient for writing thereby causing quality of writing to decrease compared to stone tablets. Followed by the evilness of books, newspapers, radio, tv, internet and so forth. And always these fringe opinions that something is outright bad/good gives away to more nuanced views.

I wonder if these people who object to using the cloud, object to using 3rd party data centres outright? Or accept that you don't have to build the physical premises where you put the compute, or do you have to own that too? If you don't have to own that, why not? Since it would seem a difficult position to at same time argue you can't use cloud because of lack of control, but you can use 3rd party data centres, now you're still lacking control on many types of outages.

If we need to own everything, where does it end? What can we get from 3rd parties? NAND gate? Or can we at least assume we don't have to build hydrogen atoms? That we get hydrogen atoms from elsewhere and start from that? Why is it that always the objection is something contemporary but the rest of the stack is fine to be provided by a 3rd party? If you believe you're living in a special period of time, where there is fundamental change to this, your position is statistically weak.

Yep, agree with all this. As I've said many times before, classic telco is no longer a model the way it used to be, and I hope that rather than fight content the way we've been doing for the past 20 years - and failing dismally - we can use this opportunity to actually work together and stay relevant, FWIW. The tides are shifting, and going against the wind has continuously worked against us. Personally, I welcome content getting involved in back-end infrastructure. It may be bitter taste for classic telco, but it significantly improves the opportunity to connect more people, more affordably. Can't argue with that. Mark.

I do disagree, if I understood the argument right. If the argument is 'cloud makes no business sense to anyone'.

That wasn't the argument I intended to make, but I see how it could have been interpreted that way. There are absolutely a ton of use cases where cloud usage makes absolute sense, from both business and technical perspectives. I do not agree that 'nobody should be using cloud providers'. There is a place for its use, and it can be very beneficial when used strategically and thoughtfully. However, as we see every time a major cloud provider has an outage, lots of people don't use it strategically or thoughtfully, and that's really my point. Randy said it most succinctly I think. Cloud providers don't reduce your risk, they just have different risks to consider. On Thu, Jan 27, 2022 at 2:32 AM Saku Ytti <saku@ytti.fi> wrote:

On Thu, 27 Jan 2022 at 09:16, Mark Tinka <mark@tinka.africa> wrote:

...

...

Saving 12 months of opex $ sounds great, except when you lose 18 months of opex $ in 2 days completely outside of your ability to control.

I don't disagree.

What this does, though, is democratize access into the industry. For a simple business model that is serving a small community with a handful of eyeballs, not trying to grow forever but put food on the table, it's somewhere to start.

I do disagree, if I understood the argument right. If the argument is 'cloud makes no business sense to anyone'.

Doing the 1st server properly costs several million euros a year, since you need competent 24/7 staffing, with sick leaves, holidays (in 1st world countries where this is a thing) and attrition taken into account. Staff who can do infra, compute, storage, networking (that's 4 separate teams usually, each needing overhead for 24/7) who are comfortable with working nights.

This argument 'no one should be using x, x is a fad' happens when every new technology appears, literally people object to using paper and pen, as it's too convenient for writing thereby causing quality of writing to decrease compared to stone tablets. Followed by the evilness of books, newspapers, radio, tv, internet and so forth. And always these fringe opinions that something is outright bad/good gives away to more nuanced views.

I wonder if these people who object to using the cloud, object to using 3rd party data centres outright? Or accept that you don't have to build the physical premises where you put the compute, or do you have to own that too? If you don't have to own that, why not? Since it would seem a difficult position to at same time argue you can't use cloud because of lack of control, but you can use 3rd party data centres, now you're still lacking control on many types of outages.

If we need to own everything, where does it end? What can we get from 3rd parties? NAND gate? Or can we at least assume we don't have to build hydrogen atoms? That we get hydrogen atoms from elsewhere and start from that? Why is it that always the objection is something contemporary but the rest of the stack is fine to be provided by a 3rd party? If you believe you're living in a special period of time, where there is fundamental change to this, your position is statistically weak.

-- ++ytti

On 1/27/22 14:43, Mike Hammett wrote:

Cloud-hosted infrastructure just doesn't work reliably. Too many points of failure along the way.

If that were true, the "Internet" (what users define as the Internet) would be down more often than not. For the price, I think there is sufficient reliability in the clouds operated by the big 3, that I'd sleep relatively okay at night if I had to use them to launch my product. Mark.

In Andreessen Horowitz's words: “you’re crazy if you don’t start in the cloud; you’re crazy if you stay on it" On 1/27/22 15:54, Michael Thomas wrote:

On 1/26/22 11:11 PM, Mark Tinka wrote:

...

On 1/26/22 17:10, Tom Beecher wrote:

...

Those folks also tend to learn hard lessons about what happens when the Magic Cloud provider fails in a way that isn't possible to anticipate because it's all black box.

Saving 12 months of opex $ sounds great, except when you lose 18 months of opex $ in 2 days completely outside of your ability to control.

I don't disagree.

What this does, though, is democratize access into the industry. For a simple business model that is serving a small community with a handful of eyeballs, not trying to grow forever but put food on the table, it's somewhere to start.

Didn't Netflix for the longest time run on AWS? I imagine if I were talking to a VC these days and said the first thing I was going to do is rack up a bunch of servers, I'd get laughed at. Cloud makes sense until it doesn't make sense. Just like everything else.

Mike

There's a big difference between a website (admittedly a complex one) and a mobile core. ----- Mike Hammett Intelligent Computing Solutions Midwest Internet Exchange The Brothers WISP ----- Original Message ----- From: "Michael Thomas" <mike@mtcc.com> To: nanog@nanog.org Sent: Thursday, January 27, 2022 3:54:57 PM Subject: Re: What do you think about the "cloudification" of mobile? On 1/26/22 11:11 PM, Mark Tinka wrote:

On 1/26/22 17:10, Tom Beecher wrote:

...

Those folks also tend to learn hard lessons about what happens when the Magic Cloud provider fails in a way that isn't possible to anticipate because it's all black box.

Saving 12 months of opex $ sounds great, except when you lose 18 months of opex $ in 2 days completely outside of your ability to control.

I don't disagree.

What this does, though, is democratize access into the industry. For a simple business model that is serving a small community with a handful of eyeballs, not trying to grow forever but put food on the table, it's somewhere to start.

Didn't Netflix for the longest time run on AWS? I imagine if I were talking to a VC these days and said the first thing I was going to do is rack up a bunch of servers, I'd get laughed at. Cloud makes sense until it doesn't make sense. Just like everything else. Mike

On 1/28/22 15:22, Josh Baird wrote:

I think Netflix's usage of AWS is being understated here.

My understanding is that the user profiles and library listings are held with AWS, but that the actual video is on their OCA's. I could be wrong... Mark.

I also think the complexities, requirements, tolerances, etc. of an EPC are also being understated in the thread. The difference being is that I am aware (and stated as such) that I'm understating Netflix's usage. The other side doesn't know how particular EPCs can be. ----- Mike Hammett Intelligent Computing Solutions Midwest Internet Exchange The Brothers WISP ----- Original Message ----- From: "Josh Baird" <joshbaird@gmail.com> To: "Mike Hammett" <nanog@ics-il.net> Cc: "Michael Thomas" <mike@mtcc.com>, "nanog group" <nanog@nanog.org> Sent: Friday, January 28, 2022 7:22:50 AM Subject: Re: What do you think about the "cloudification" of mobile? I think Netflix's usage of AWS is being understated here. On Fri, Jan 28, 2022 at 6:29 AM Mike Hammett < nanog@ics-il.net > wrote: There's a big difference between a website (admittedly a complex one) and a mobile core. ----- Mike Hammett Intelligent Computing Solutions Midwest Internet Exchange The Brothers WISP From: "Michael Thomas" < mike@mtcc.com > To: nanog@nanog.org Sent: Thursday, January 27, 2022 3:54:57 PM Subject: Re: What do you think about the "cloudification" of mobile? On 1/26/22 11:11 PM, Mark Tinka wrote:

On 1/26/22 17:10, Tom Beecher wrote:

...

Those folks also tend to learn hard lessons about what happens when the Magic Cloud provider fails in a way that isn't possible to anticipate because it's all black box.

Saving 12 months of opex $ sounds great, except when you lose 18 months of opex $ in 2 days completely outside of your ability to control.

I don't disagree.

What this does, though, is democratize access into the industry. For a simple business model that is serving a small community with a handful of eyeballs, not trying to grow forever but put food on the table, it's somewhere to start.

Didn't Netflix for the longest time run on AWS? I imagine if I were talking to a VC these days and said the first thing I was going to do is rack up a bunch of servers, I'd get laughed at. Cloud makes sense until it doesn't make sense. Just like everything else. Mike

On 1/28/22 13:28, Mike Hammett wrote:

There's a big difference between a website (admittedly a complex one) and a mobile core.

Word is it hasn't been smooth-sailing, but Amazon are pushing on. Failing and improving is in their DNA, so I'm sure everyday, they are one step closer to the promised land. Affirmed Networks, I believe, was their first engagement re: vEPC. Mark.