Watch on YouTube – The Internet Report – Ep. 19: Aug 10 – Aug 16, 2020

This is the Internet Report, where we uncover what’s working and what’s breaking on the Internet—and why. On this week’s episode, Archana and I cover recent headlines concerning social media platform, TikTok, and the gaming provider, Epic. TikTok appears to have gained some additional time (now 90 days) before the US government will enforce its ban on the service. Gaming provider, Epic, recently made news when its game Fortnight was removed from Apple’s App Store and Google’s Play Store for violating their Terms of Service. Epic was quick to file a lawsuit claiming the tech giants were in violation of anti-competition laws. The outcome of this case will be one to watch, and can have far-reaching impacts for developers. Next up, we speak with William Collins, Lead Cloud Architect at a Fortune 100 company, about cloud connectivity, on-ramp services and the difference between the “Big 3” on-ramp services.

Find us on:

Finally, don’t forget to leave a comment here or on Twitter, tagging @ThousandEyes and using the hashtag #TheInternetReport.

Catch up on past episodes of The Internet Report here.

Listen on Transistor – The Internet Report – Ep. 19: Aug 10 – Aug 16, 2020
ThousandEyes T-shirt Offer

Follow Along with the Transcript

Angelique Medina:
This is The Internet Report, where we uncover what’s working and what’s breaking on the Internet and why. Last week was a bit of a slow week, but there were some interesting developments in terms of applications that have been in the news recently, including TikTok and Fortnite. We also have a really interesting interview with a lead cloud architect out of a Fortune 100 company, and he’s going to talk a little bit about cloud connectivity and some of the services they use. And so, it’s going to be really interesting to get some of his insight.

Angelique Medina:
But before we get to that, we saw last week that Fortnite was pulled from Apple and Google stores for violating other terms of usage and trying to make an end run around the collection of payments for in-app purchases. And that’s really interesting because that effectively kicked off a chain of events where they got pulled from the store, and then they immediately filed an anti-trust lawsuit in court, so this was something that they clearly had planned. It’s interesting because they allude to the “Apple tax,” if you will, which takes a 30% cut of all in-app purchases, which is pretty, pretty high. It was surprising to me. I knew that they took a pretty sizable cut, but I didn’t realize it was 30%. And they’re effectively alleging that this is a monopoly that Google and Apple have created where no one can really offer any kind of application, unless they’re really doing everything that Apple and Google ask of them.

Angelique Medina:
It was interesting because Microsoft also had a statement about this in terms of how Apple treats certain types of applications like gaming applications, for example.

Archana Kesavan:
As not their priority. I think in terms of the violation itself, how they orchestrated it is, it essentially give users another way in to make in-app purchases, so something that bypasses the App Store itself. So the 30% cut that Apple takes happens only if you go through the store, and if you don’t go do that, then that’s the violation that kicked Fortnite out of the App Store, and it looks like Google as well, right?

Angelique Medina:
Yeah. Google as well. So they shut that down pretty quickly, but it seems like there’s maybe a broader issue at play, even besides this cut that they’re taking from the application developers. It’s also alleged by Microsoft, that they unfairly treat gaming applications, and treat that somehow differently from other apps that are offered on their …

Archana Kesavan:
On the App Store.

Angelique Medina:
Yeah.

Archana Kesavan:
I think it’s going to be interesting to see how this plays out, not just from a Fortnite perspective, but if people start following it, other folks who go through the App Store start following it, then it’s going to change how the App Store ecosystem works. So it’s going to be pretty significant to just see how this plays out and the end result of this.

Angelique Medina:
Yeah. And It reminds me of a similar situation where, for example, artists were rebelling against how they were compensated through streaming services like Spotify and so on. The developers are… I mean, people want the content, they want the applications themselves, so if they’re not available through the App Store, then that’s going to be a problem for both Apple and Google. And so, it’ll be interesting to see how collectively the power of the app developers potentially increases as a result of maybe other folks joining this. It remains to be seen, but certainly an interesting development and something that we’ll continue to look at and see how it goes.

Archana Kesavan:
And then talking about apps, the other newsworthy headline for this week is TikTok. Banning TikTok from the US, that deadline has shifted and extended to about 90 days. So, that takes this all the way up to November.

Angelique Medina:
Yeah, that’s right. And so, that certainly opens the window for Microsoft to continue their discussions. And there’s been talk, not only of them potentially taking over their business in the US, but also in other English speaking countries, like the UK, and New Zealand, and Australia and so on. So, something to watch for sure. So with that, I think we’re going to go ahead and transition to talking about cloud deployments. This is a really interesting discussion that you had with our guest William Collins.

Archana Kesavan:
Right. Right. I actually bumped into William from a Packet Pushers conversation where he was talking about multi-cloud and on-ramp services. And William has been really involved in terms of cloud design for a leading Fortune 100 company. So what we’re talking about today is in terms of the evolution of cloud connectivity, what enterprises started off with, and how they slowly move on to an on-ramp service. And then for me, the most interesting piece was around the difference between these on-ramps services that cloud providers offer, because we’ve been doing some kind of comparison from a performance perspective across cloud providers. It was really interesting to listen to his take. So stay tuned for that on The Expert Spotlight.

Archana Kesavan:
Oh, welcome to The Expert Spotlight Spread spotlight. This week, we actually have William Collins. William is the lead cloud architect that works for a Fortune 100 company. And over his career, he has designed large scale networks, driven modernizations for cloud adoption, and looks to constantly optimize through good design practices in automation. His most recent work has been around hybrid and multi-cloud focused, in AWS, Azure, Google Cloud, Equinix. But outside of a stack, his time has been spent with family, woodworking, ice hockey and classical guitar, so. So William, thank you so much for being on the show today.

William Collins:
Thanks for having me.

Archana Kesavan:
So I think one of the things we wanted to focus this discussion around, William, is around your background, talk about multicloud, but also talk about connectivity into the cloud, and then get into some details there. But to kick this session off, how do you think about evolution when an enterprise is migrating to the cloud specifically from a connectivity perspective?

William Collins:
Sure. That’s a good question. So nobody started on multi-cloud, right? We all started out, usually it was AWS, and oftentimes it was just the VPN from the data center. And that’s when we started figuring out, “Okay, this cloud thing, how does it enable our business? The overall effects it has on application delivery, and how can we integrate it in with our business model and what we’re doing?” So VPNs are easy. Most enterprises have some process in place to spin up a business to business, or land to land VPN for a partner. And, off we go, and then, at some point you realize, “Hey, what are these Direct Connects, Express Routes, Interconnects, all these really fast private means of connectivity?”

William Collins:
So then, what an enterprise will do is look at a lot of different partner connections. So AT&T, NetBond, CenturyLink Cloud Connect, even Equinix Cloud Exchange. So what that enables you to do is, instead of having these single VPNs maybe from multiple data centers, it’ll allow you to piggyback off your existing service with those providers. So, say you leverage AT&T for MPLS and you want to extend it to AWS Direct Connect. So you can basically say, “Okay, AT&T, I want to advertise these routes over this connection to AWS, and set up your A side and B side VLAN, turn on BGP and, off you go.”

William Collins:
And then, when you start looking at really heavy performance considerations, and really looking at your design over time and how you want to scale, and maybe you start looking at multicloud, and then, the co-lo discussion will come up and that’s where you have these one for one Direct Connects and ExpressRoutes and Direct Connect. So, instead of having a partner in the middle, you’re connecting with bare metal and you’re going to terminate it on your own switches and set up your own routing and your own policy to facilitate all this.

Archana Kesavan:
How does the evolution happen in a building that co-lo and deciding where to pick that connection and get all of them meshed of sorts?

William Collins:
Yeah. That’s a really good question. So the way that I’ve seen it happen pretty much everywhere I’ve put in design work with this kind of stuff is, usually the application developers will choose what region they initially want to go to. And usually, they’re going to choose the one that’s closer to them or closer to the business. So oftentimes, what you see is, “Okay. Well, that’s where we’re going to maybe put our co-los. We want a really close proximity to the cloud regions we’re using,” So that’s how that happens. And then, you move off from, instead of just availability zones, and really looking at redundancy route, then you want to look at multi region from that point, and then spread out that way.

Archana Kesavan:
Got it. Cool. On these on-ramp services and this one to one connectivity that you were talking about. The big three have their own variety of it, you have the ExpressRoute, you have Direct Connect and Google Interconnect. From your experience, while in theory, these are all on-ramp services. What’s the difference between these three?

William Collins:
Sure. So when you think about the Direct Connect, ExpressRoute, Interconnect and Fast Connect, these are usually layer two vehicles for connecting to that cloud provider. So oftentimes, what you’re going to have with those services at that level is, you’re going to have just the, how you initially set off the process to connect one. And oftentimes, I know that for a while there, AWS had physical Direct Connects available for some time. And when Azure really got into the space, Microsoft did with ExpressRoute for a while through an Equinix, you couldn’t do a one to one direct connection with Microsoft. You had to use a partner connection of some sort, like Cloud Exchange from Equinix or Megaport. And even with Google Interconnect, I think they had physical Direct Connects available from the get go when they launched Interconnect.

William Collins:
So thinking of that, you have sort of a different look and feel there because, on one side you have physical connections, and then on the other side, you’re bringing in a whole new partner connection in between, so it can get complicated. But since then, Microsoft does have physical ExpressRoutes now, so that’s changed. And then, route advertisements too, some routing policies. So they’ll have different numbers to how many routes you can actually advertise up into their services, which can bite you sometimes, that’s bit me in the past actually. And to give you an idea of what happens when you advertise more routes up than you should, the Direct Connect goes down. Yeah, you go hard down. So you want to make sure that you’re on top of those small details that can come back and get you.

Archana Kesavan:
Oh, one of the things again… Well, thinking about connectivity, we’re starting to see these private connectivity links come up, for instance, Azure’s private link. How does it contrast and compare with an express route? And what are the use cases you would think about for using one versus the other?

William Collins:
With an Azure private link. Azure private link, you can think about, as of today, it sort of compliments your existing connectivity over ExpressRoute. And basically, what it’s going to allow you to do is, within an Azure VNet, you have your own RFC 1918 space for holding that, maybe /20, /24, whatever you have. So what private link will allow you to do is privately interface with Microsoft different path services, their managed services that you’re probably already using, over that private address space. There’s other ways that organizations have gone to make sure, or to try to get private connectivity, or at least limit the footprint and make it as private as they can, but this makes it easy because it actually takes IP addresses straight out of your VNet and allows you to interface that way.

Archana Kesavan:
Got it, and also …

William Collins:
So basically, their accessing dedicated path services are integrated directly with their VNet. It would be like using ExpressRoute private tunnel in some sort of VPN, removing Internet traffic completely, but still being able to reach in directly to that path service privately.

Archana Kesavan:
Right. Right. It’s a common trend in this whole situation of bypassing the Internet and these cloud providers trying to monetize their own backbone. We saw AWS Global Accelerator, we’re now talking about Azure private link. What’s your thought about this trend?

William Collins:
Yeah. Well, even if you look at SD-WAN. SD-WAN comes on the market for us, networking folks, to say, “Hey, we’re going to give you an overlay that’s transport agnostic. You don’t have to do MPLS circuits everywhere,” but what the enterprise is probably going to do at the get go is, run SD-WAN on top of MPLS. So, very risk aversion at its finest. And then, if you look at public cloud, the premise of public cloud is to use the Internet. For a while there, these private connections, Direct Connects ExpressRoutes, they weren’t there. A lot of this stuff was brought on from enterprise customer demand. So I think, at the end of the day, a lot of enterprises are going to try and force everything they can, and connect over private connectivity, if at all possible, using their own RFC 1918 space.

William Collins:
And part of that too, is the demand and the driver. So a lot of big businesses, they have acquired a lot of data and a lot of applications over the years that exist in their own on-premise at data centers. So when you get into single cloud, multi-cloud, all this stuff at some point, it’s got to talk to each other. So in terms of migrating applications, a tiered hybrid approach to applications where, maybe a web front end exists in the cloud, but you still have reusable services back in the data center of lots of things you use. Well, what’s the easy way to do that? Is having that private connection in between.

Archana Kesavan:
Thank you so much for your time and walking us through this.

William Collins:
Absolutely. Thanks for having me on.

Angelique Medina:
That was a really interesting discussion, Archana. It’s going to be interesting to see how the network services that are offered by the cloud providers evolved, because it seems like every year, there’s some new offering that comes out, and it’s definitely …

Archana Kesavan:
I know. They just find new ways to monetize their backbone, right? They’ve spent so many years and a lot of dollars building that infrastructure. It makes sense for them to monetize that. But I think what’s interesting for us to keep in touch is to identify how these services are different from each other. They’re just so many of them popping up, and the marketing wrapper around it sometimes obscures why you use one versus the other. So I think that’s an interesting topic just to get into.

Angelique Medina:
Absolutely. How they differ from one another. And even if they’re needed within the context of just the public Internet, because there’s a lot of marketing hype around this. Cloudflare launched their own Interconnect, and they messaged a lot around how you can avoid exposure to the public Internet. So I think that leads to the obvious question of, do you really need to be protected from the Internet? Maybe you do, maybe you don’t, but that’s a fair question.

Archana Kesavan:
No, totally. And I think the fear around the Internet takes people down the path of, these services do make an impact and there are performance benefits. I guess it comes down to, what is the return on investment on this performance benefit, right? And we saw that with the AWS Global Accelerator in our research last year, that performance does vary. So question is, if you’re investing in a service, just baselining to understand what you can get out of that service, I guess, becomes important for the enterprise.

Angelique Medina:
Absolutely. All right. Well, that’s our show. So don’t forget to subscribe and follow us on Twitter. And as always, if you have questions, or feedback, or guests, or topics you would like to see covered on the show, feel free to drop us a note at InternetReport@thousandeyes.com. And if you do subscribe, you can claim a free T-shirt. Again, just send an email to InternetReport@thousandeyes.com with your address and T-shirt size, and we’ll get that shipped right over for you. All right, till next time.

Subscribe to the Internet and Cloud Intelligence Blog!
Subscribe
Back to ThousandEyes Blog