Digital Edge Strategy and Why Enterprises Should Care - Podcast transcript

On October 12th, 2017, top cloud influencer, David Linthicum and I got together to discuss the importance of your company's digital edge strategy and how Interconnection can help you optimize this strategy from a latency, security and cost perspective. The audio discussion can be found at the Cloud Weekly Podcast (subscribe here) and on the Cloud Technology Partners web site.

David: Anything new going on in (the world of Equinix) since we last talked?

James: Oh yeah, it's been fantastic. We've made a couple acquisitions which have grown our footprint globally, so we can now satisfy customers needs to be in pretty much every region of the globe where they are doing business. And we recently published a new set of research we call the Global Interconnection Index which really shows the growth of the private connectivity that companies are needing to do to really help them deliver better solutions with lower latency, violation-protections so they don't have the hacking problems you run into when sharing information over the public Internet, and significant cost savings as well.

David: So tell us some of the details behind what your company does, because there's a lot of companies in the same space but you guys are bigger than most and provide a lot more services than your competitors. So fill us in.

James: Equinix is the largest carrier-neutral and cloud-neutral provider of Interconnection. And we do this through a series of Data Centers - the largest footprint of global data centers of all the providers out there. And we do this through a marketplace that allows you to do connectivity to carriers (to do, say an AT&T connection to Verizon), but also to connect to any of the major public clouds from almost any major metros across the world, as well as connectivity to the leading SaaS providers, most of the leading financial services firms (especially if you are moving down the path to doing eCommerce in your digital strategy) and you can also do connections to other parts of your ecosystem. So you can really build out rich, robust digital applications you can deliver with low latency, good security and full compliance with the law, all across the globe.

David: One of the things my clients often ask is if we can't migrate everything to the cloud or everything into SaaS or look for platform analogs in the cloud we have to maintain existing solutions and existing data centers. You guys provide a solutions right? The ability to link up to the various cloud providers, you're able to manage the connections between the systems, able to outsource the stuff that enterprise data centers are finding very difficult and expensive to maintain.

James: That is exactly right. And we work very much with companies like CloudTP and others who can help manage these relationships for you and make sure they are properly set up and configured. Something else we recently did is bring forth the Interconnection Oriented Architecture Knowledge Base, which, similar to GitHub, is an open community where everybody can come in and talk about the best practices for their hybrid-, multi-cloud implementations - how they secured it, how they delivered great performance and so on, and share that with the whole community.

David: One of the things I hear out there is this whole Digital Edge strategy and Interconnectivity specifically. You guys seem to be working on stuff like that. That's fascinating to me. So kind of fill us in on what that is and what it means to your customers and how that's going to grow in the future and why enterprises should care about that.

James: It's a great topic and one we cover with our clients every single day. Because as everyone knows, almost every industry is moving to digital. You can't be satisfied with traditional products. If you think about the manufacturing space, every product you build now has two new aspects: it better have sensors and you better be gathering that information and using machine learning and AI to feed information back to the buyer of the product about what is going on, what's the best use, am I safe, what changes I need to make, and so forth. And that applies not only to autonomous cars (that case pretty straight forward) but certainly we are seeing this happen with fitness and health gear, but also elevators and buildings and with the military. It's pretty much happening all over the place. And exactly to those points, when you sell these goods, or you are an insurance company who is going to insure those goods, or a healthcare company that has to pull insights from those goods, you can't do that if you have to drag that data all the way back to your headquarters to do the machine learning in your data warehouse. You've got to create as much insight and as close to real-time as possible at those devices or in the metros where those devices are located. Which means you have to have a Digital Edge strategy that really does collect, analyst and deliver insights in every single metro your company does business in. And that's why you have to really be thinking about, "What is my Digital Edge? How many metros am I in? Where are my customers? And how do I satisfy them?"
One good example of this: I was just in Europe and started having some knee problems. And our company has a healthcare application (for employees) that allows me to take a photo of an injury, send it to a doctor, have him check it out and say, "I think you're ok" or "Oh! Go to the emergency room right away!"
That application had such horrible latency in London that I simply couldn't use it. And they didn't think they needed to be in London because they sell to US companies only. Well, guess what? Your Digital Edge is wherever Americans go on vacation.

David: Yeah, that's when you're going to need the help. You're only going to take a picture of your injured leg when you are out of the country and can't get down to the emergency room. It's confusing but ultimately, that's the value they can bring. It's not the easy stuff; it's the hard stuff.
I was talking to a reporter the other day about what's going to change over the next 5 years. We're going to move into this access to perfect, realtime information; the ability to see anything we want to see, as its happening, and the ability to make instant analytical calculations based on the information as well as understand trends and do predictive analytics. And really get into the part we've been trying to do for the last 30 years - zero-latency enterprises and frictionless commerce, but we never really got there - I wrote 3 books about it. And still it seems to be a big issue.
So are we going to reach a saturation point where we just have too much information coming in from these sensors and from these devices? Will we need to have aggregation layers to deal with it. Do we need to figure out how to manage this information? We have clients, especially doing IoT stuff on the factor floor, where they have petabytes of information coming into their database. And obviously that's too much information that they're aggregating from. But there needs to be a strategy for how we're analyzing the data and people typically aren't thinking about that. Their gathering everything, not necessarily understanding what the information means. Comments on that?

James: You're exactly right. This is what people need to think about: What data do you keep and in what locations do you keep it? What I hear a lot of companies say is, "OK great, but this manufacturing equipment that we sell is generating a TB of data per day. I can't afford to put that kind of storage into my devices and raise the price of the device ridiculously. And I don't know how I'll do the analytics. So what I'm going to do is only capture the alerts. And I'm going to send them and do something with that data. And maybe I'll send the data a couple minutes before an alert or after it and will send that back to HQ and we'll do the analysis."
That's great if you have 100% confidence the alerts catch all the problems. If you don't have that level of confidence you are going to miss a lot of potential opportunities and you have dumped that data. So if I'm going to collect that data, where should I collect it? How much? And should I do some analysis that looks at the data and identifies patterns that I need to look at and then says, "OK, before I was only keeping 5% of the data when alerts came out. Now that I have done this kind of analysis, now I can see patterns that show trends toward things where an alert might be needed. That's only 30% of the data. Great let's keep that and move that 30% over and then combine it with data from all the other data from devices in this same metro or country, and then we can aggregate even more data from the cloud that we store in Google or Salesforce about our customers, so we can use that to go back to our customer with recommendations of what they should be doing. All of this creates more data, but you don't need to store it all in the device or bring it all back to the data warehouse.

You have to optimize what amount should be in the device, stored at the edge or brought back. And in each of those locations, what other data should we Interconnect to? Because, for example, if I'm doing sensors in a car, it's great that the car can analyze that the tire is getting low or its temperature is going way up. But the car alone has no idea if other cars in the same city are experiencing the same things. Or that what this is, is a tornado coming down. That's what I need information from the Metro for and that's what I need Interconnectivity to The Weather Company and other places who can pull this data together, so I can go back to the car owner and say, "We noticed where you are. There's a tornado coming your way. Go left and get out of here."

David: Yeah, I wrote an article for IEEE IoT Journal last year that talked about the tiering of data and how to analyze information. Because we ran into the same thing: People have a tendency to want to drink from the firehose when they don't really need to. It's a matter of understanding what's meaningful in terms of info coming off these sensors. We have to remember that sensor providers don't necessarily understand your application, in terms of how you're going to leverage their information. So, via their APIs, they're going to send you everything you ask for. But they are going to provide you with selective access to certain information, even storing and analyzing some of the information on the device. But when to transmit it back to intermediate databases, cloud-based DBs, etc., how to deal with latency issues and things like that - we're just figuring this out right now. It's funny, we have the tools and how to do this but not necessarily how to tier the data.

One piece of information you brought to my attention - the Global Interconnection Index forecasts that overall bandwidth of private connections at IT exchange points will outgrow the Internet by 2x and Internet bandwidth by 6x. and MPLS WAN connectivity by 10x! I find that surprising. Can you elaborate on that a bit.

James: Sure. Happy to. Just as we were talking about, if you're thinking about an IoT model and want to move information from here to here, that's important information that you want to be secure. And you really don't want to move it over an Internet link or you may want to move subsets of that info that has personally identifiable records over another way. The best way, and the growth that you are seeing there, are these private connections. And what's increasingly happening is that you can move these private connections from your sensor device into the Amazon IoT Suite and from there want to move it to a repository you keep off of the cloud, for whatever reason that you need to keep it separate from the cloud. If you know that in the metro you are collecting all this data, that Amazon has a location there and you have a cabinet inside an Equinix colo facility, you can actually set up a private LAN-speed connection between Amazon and your cabinet. That allows you to move enormous amounts of data and not be worried about latency, the cost of moving the data, and you certainly have superior security when it is a private 1:1 connection. And as you think about more and more companies moving to a Digital Edge model they're going to be doing a lot more Interconnecting between their cloud, SaaS and other partners in those metros. And if I can do all of that over LAN connections, then I'm not worried about security and the bandwidth. That's what really is the driver.
You're still going to use Internet connections to go to the edge because you aren't going to do a private link to every single customer in that metro. And you're still going to need WAN to go from metro to metro and back to your HQs. But because you can do so much more analysis and Interconnection, right there in each metro you do business in,  that's what causes that traffic shift that is articulated in the forecast.

David: Got it. So how are enterprises getting around this right now? Are they using your platform to do this? Are they failing at it. Are they trying to upgrade their network infrastructure? Trying to become a data center provider. What's working and what's not?

James: So, obviously this is a forecast that isn't Equinix-specific. Everyone in the colo and Interconnection business are able to create these solutions. So what we're finding as enterprises focus more and more on building out there Digital Edge, they don't really want to buy or build out there own data centers. They would rather put it somewhere fast and easy. Same reason they go out and use the cloud. But as we were talking about before, some data you just can't put in the cloud or you don't trust to put it in the cloud because you have partnerships with certain governments, for example.
So you really do need to have a hybrid play as you build out all these metros.

And you really do have to rethink your architecture because, as you know -- we've seen this for years in enterprises -- there's been an old model of a primary data center plus a DR site and then you use WAN links to all your offices and places of business. That model doesn't work very well anymore because all the traffic had to come back over the WAN to HQ in that type of model. It's like the VPN model that employees get frustrated by all the time: "I'm connecting to Salesforce, who has a data center right here. What do you mean I have to VPN back to our HQ in London? This is ridiculous." We see this model a lot. And what a lot of companies are starting to realize is that if they move to a Digital Edge model, and away from the WAN-to-everywhere model, just the cost savings from moving from a WAN-centric model to an Interconnection model justifies the investment in the move. 

David: So where can I go to get best practices for this? You guys and the industry are certainly promoting it but where can I go for open discussion and talk to others who are doing this and get unbiased approaches, and things like that?

James: You're exactly right and your question is correct. It's not just simply that I'm going to go from a WAN connection to an Interconnection and just move my data from this location to that one. You're probably going to need to do some rearchitecting of your app. So there are a lot of changes you need to make to take it there. That's why we created the Interconnection Oriented Architecture Knowledge Base (ioakb.com) which is an open community where the best practices of moving to an IOA model are published. And where you can engage with experts - not just by us - but other enterprises, partners, service providers, MSPs and other companies who are building out these practices. It's a great place to go that is cloud-neutral, carrier-neutral and even Equinix-neutral, where you can find out what are the best practices for moving to this new Digital Edge model.

David: Yeah, I love doing that because there's always an upside and a downside and you learn from the mistakes of others, so to speak. :)
I've built many architectures on the backs of failures of other people. :)
And enterprises aren't doing as best a job as they can in understanding the context of something. They get enamored with the technology and the approach and it really becomes the center of the focus instead of their requirements and how to fit them into a particular solution. At the end of the day we're trying to configure this correctly, get the right enabling technologies in place - whether it's cloud or not. But enterprises have a tendency to get lost around this stuff, especially the new stuff. We're diving into this now and see a lot of mistakes being made.

So what would your elevator guidance be for someone investing in a Digital Edge strategy? -- What do I do, how do I get started and what are the two most important things I need to think about.
And BTW, I just hit the 16th floor.

James: Nice :) First and foremost, understand what metros you should be in and second what do you need to do to your app to move it into a metro-centric model. And are you doing this to address latency? Are you doing it in a way that can be solved simply by caching more content, using Akamai? Typically that's not good enough anymore. Because customers want real time interaction with your app and so, what part's of the app need to be local? Because you can totally over-rotate and move way too much and get into a situation where it is more expensive than before. You don't want to do that.
Then what's easy, once you are in the locations that matter, you can go to places like the Equinix Marketplace, to see which of your partners are located in same metro that you can Interconnect to. And through our portal you can just "click, click, click" and set up those connections. And if you are building out a hybrid application, say using Azure Machine Learning and Google and Salesforce for some data you are pulling in, knowing you can make those connections, right there, and that they will be private and secure, makes it super easy to focus on:

• Did I design the app right
• Am I giving the right customer experience
• And if so, great, let's move forward.

David: Yeah, check it out. It's a good concept and a lot of enterprises are starting to move in that direction these days. It's always good to build your stuff on the knowledge of other people's experience.

Please pick up a copy of my book Cloud Computing & SOA Convergence, read my blog on Infoworld also check out my new cloud computing courses on Lynda.com. And be sure to go www.CloudTP.com/Doppler for current cloud news and how to connect to us directly.
You guys take good care. Goodbye.