The downside of it is, at the end of the day, that the value in any computing system is not so much in the hardware per se; it’s in the data that’s the basis of how the system works. And because of the realities of working with data in a distributed way, globalization that is intended to more fully enfranchise data wherever it might be introduces a range of architectural implementation and legal complexities that can’t be discounted.
So, cloud and globalization can go together -- but it dramatically increases the need for smart and forward-thinking approaches to imagining, and then ultimately realizing, how those two go together, and what hybrid architecture is going to be required to make it work.
Gardner: If you need to then focus more on the data issues -- such as compliance, regulation, and data sovereignty -- how is that different from taking an applications-centric view of things?
Learn More About
Solutions From HPE
Burris: Most companies have historically taken an infrastructure-centric approach to things. They start by saying, “Where do I have infrastructure, where do I have servers and storage, do I have the capacity for this group of resources, and can I bring the applications up here?” And if the answer is yes, then you try to ultimately economize on those assets and build the application there.
That runs into problems when we start thinking about privacy, and in ensuring that local markets and local approaches to intellectual property management can be accommodated.
But the issue is more than just things like the General Data Protection Regulation (GDPR) in Europe, which is a series of regulations in the European Union (EU) that are intended to protect consumers from what the EU would regard as inappropriate leveraging and derivative use of their data.
It can be extremely expensive and sometimes impossible to even conceive of a global cloud strategy where the service is being consumed a few thousand miles away from where the data resides, if there is any dependency on time and how that works.
Ultimately, the globe is a big place. It’s 12,000 miles or so from point A to the farthest point B, and physics still matters. So, the first thing we have to worry about when we think about globalization is the cost of latency and the cost of bandwidth of moving data -- either small or very large -- across different regions. It can be extremely expensive and sometimes impossible to even conceive of a global cloud strategy where the service is being consumed a few thousand miles away from where the data resides, if there is any dependency on time and how that works.
So, the issues of privacy, the issues of local control of data are also very important, but the first and most important consideration for every business needs to be: Can I actually run the application where I want to, given the realities of latency? And number two: Can I run the application where I want to given the realities of bandwidth? This issue can completely overwhelm all other costs for data-rich, data-intensive applications over distance.
Gardner: As you are factoring your architecture, you need to take these local considerations into account, particularly when you are factoring costs. If you have to do some heavy lifting and make your bandwidth capable, it might be better to have a local closet-sized data center, because they are small and efficient these days, and you can stick with a private cloud or on-premises approach. At the least, you should factor the economic basis for comparison, with all these other variables you brought up.
Burris: That’s correct. In fact, we call them “edge centers.” For example, if the application features any familiarity with Internet of Things (IoT), then there will likely be some degree of latency considerations obtained, and the cost of doing a round trip message over a few thousand miles can be pretty significant when we consider the total cost of how fast computing can be done these days.
The first consideration is what are the impacts of latency for an application workload like IoT and is that intending to drive more automation into the system? Imagine, if you will, the businessperson who says, “I would like to enter into a new market expand my presence in the market in a cost-effective way. And to do that, I want to have the system be more fully automated as it serves that particular market or that particular group of customers. And perhaps it’s something that looks more process manufacturing-oriented or something along those lines that has IoT capabilities.”
The goal is to bring in the technology in a way that does not explode the administration, management, and labor cost associated with the implementation.
The goal, therefore, is to bring in the technology in a way that does not explode the administration, managements, and labor cost associated with the implementation.
The other way you are going to do that is if you do introduce a fair amount of automation and if, in fact, that automation is capable of operating within the time constraints required by those automated moments, as we call them.
If the round-trip cost of moving the data from a remote global location back to somewhere in North America -- independent of whether it’s legal or not – comes at a cost that exceeds the automation moment, then you just flat out can’t do it. Now, that is the most obvious and stringent consideration.
On top of that, these moments of automation necessitate significant amounts of data being generated and captured. We have done model studies where, for example, the cost of moving data out of a small wind farm can be 10 times as expensive. It can cost hundreds of thousands of dollars a year to do relatively simple and straightforward types of data analysis on the performance of that wind farm.
Process locally, act globally
It’s a lot better to have a local presence that can handle local processing requirements against models that are operating against locally derived data or locally generated data, and let that work be automated with only periodic visibility into how the overall system is working closely. And that’s where a lot of this kind of on-premise hybrid cloud thinking is starting.
It gets more complex than in a relatively simple environment like a wind farm, but nonetheless, the amount of processing power that’s necessary to run some of those kinds of models can get pretty significant. We are going to see a lot more of this kind of analytic work be pushed directly down to the devices themselves. So, the Sense, Infer, and Act loop will occur very, very closely in some of those devices. We will try to keep as much of that data as we can local.
But there are always going to be circumstances when we have to generate visibility across devices, we have to do local training of the data, we have to test the data or the models that we are developing locally, and all those things start to argue for sometimes much larger classes of systems.
Gardner: It’s a fascinating subject as to what to push down the edge given that the storage cost and processing costs are down and footprint is down and what to then use the public cloud environment or Infrastructure-as-a-Service (IaaS) environment for.
But before we go into any further, Peter, tell us about yourself, and your organization, Wikibon.
Learn More About
Solutions From HPE
Burris: Wikibon is a research firm that’s affiliated with something known as TheCUBE. TheCUBE conducts about 5,000 interviews per year with thought leaders at various locations, often on-site at large conferences.
I came to Wikibon from Forrester Research, and before that I had been a part of META Group, which was purchased by Gartner. I have a longstanding history in this business. I have also worked with IT organizations, and also worked inside technology marketing in a couple of different places. So, I have been around.
Wikibon's objective is to help mid-sized to large enterprises traverse the challenges of digital transformation. Our opinion is that digital transformation actually does mean something. It's not just a set of bromides about multichannel or omnichannel or being “uberized,” or anything along those lines.
The difference between a business and a digital business is the degree to which data is used as an asset.
The difference between a business and a digital business is the degree to which data is used as an asset. In a digital business, data absolutely is used as a differentiating asset for creating and keeping customers.
We look at the challenges of what does it mean to use data differently, how to capture it differently, which is a lot of what IoT is about. We look at how to turn it into business value, which is a lot of what big data and these advanced analytics like artificial intelligence (AI), machine learning and deep learning are all about. And then finally, how to create the next generation of applications that actually act on behalf of the brand with a fair degree of autonomy, which is what we call “systems of agency” are all about. And then ultimately how cloud and historical infrastructure are going to come together and be optimized to support all those requirements.
We are looking at digital business transformation as a relatively holistic thing that includes IT leadership, business leadership, and, crucially, new classes of partnerships to ensure that the services that are required are appropriately contracted for and can be sustained as it becomes an increasing feature of any company’s value proposition. That's what we do.
Global risk and reward
Gardner: We have talked about the tension between public and private cloud in a global environment through speeds and feeds, and technology. I would like to elevate it to the issues of culture, politics and perception. Because in recent years, with offshoring and looking at intellectual property concerns in other countries, the fact is that all the major hyperscale cloud providers are US-based corporations. There is a wide ecosystem of other second tier providers, but certainly in the top tier.
Is that something that should concern people when it comes to risk to companies that are based outside of the US? What’s the level of risk when it comes to putting all your eggs in the basket of a company that's US-based?
Burris: There are two perspectives on that, but let me add one more just check on this. Alibaba clearly is one of the top-tier, and they are not based in the US and that may be one of the advantages that they have. So, I think we are starting to see some new hyperscalers emerge, and we will see whether or not one will emerge in Europe.
I had gotten into a significant argument with a group of people not too long ago on this, and I tend to think that the political environment almost guarantees that we will get some kind of scale in Europe for a major cloud provider.
If you are a US company, are you concerned about how intellectual property is treated elsewhere? Similarly, if you are a non-US company, are you concerned that the US companies are typically operating under US law, which increasingly is demanding that some of these hyperscale firms be relatively liberal, shall we say, in how they share their data with the government? This is going to be one of the key issues that influence choices of technology over the course of the next few years.
Cross-border compute concerns
We think there are three fundamental concerns that every firm is going to have to worry about.
I mentioned one, the physics of cloud computing. That includes latency and bandwidth. One computer science professor told me years ago, “Latency is the domain of God, and bandwidth is the domain of man.” We may see bandwidth costs come down over the next few years, but let's just lump those two things together because they are physical realities.
The second one, as we talked about, is the idea of privacy and the legal implications.
The third one is intellectual property control and concerns, and this is going to be an area that faces enormous change over the course of the next few years. It’s in conjunction with legal questions on contracting and business practices.
Learn More About
Solutions From HPE
From our perspective, a US firm that wants to operate in a location that features a more relaxed regime for intellectual property absolutely needs to be concerned. And the reason why they need to be concerned is data is unlike any other asset that businesses work with. Virtually every asset follows the laws of scarcity.
Money, you can put it here or you can put it there. Time, people, you can put here or you can put there. That machine can be dedicated to this kind of wire or that kind of wire.
Data is weird, because data can be copied, data can be shared. The value of data appreciates as we us it more successfully, as we integrate it and share it across multiple applications.
Scarcity is a dominant feature of how we think about generating returns on assets. Data is weird, though, because data can be copied, data can be shared. Indeed, the value of data appreciates as we use it more successfully, as we use it more completely, as we integrate it and share it across multiple applications.
And that is where the concern is, because if I have data in one location, two things could possibly happen. One is if it gets copied and stolen, and there are a lot of implications to that. And two, if there are rules and regulations in place that restrict how I can combine that data with other sources of data. That means if, for example, my customer data in Germany may not appreciate, or may not be able to generate the same types of returns as my customer data in the US.
Now, that sets aside any moral question of whether or not Germany or the US has better privacy laws and protects the consumers better. But if you are basing investments on how you can use data in the US, and presuming a similar type of approach in most other places, you are absolutely right. On the one hand, you probably aren’t going to be able to generate the total value of your data because of restrictions on its use; and number two, you have to be very careful about concerns related to data leakage and the appropriation of your data by unintended third parties.
Gardner: There is the concern about the appropriation of the data by governments, including the United States with the PATRIOT Act. And there are ways in which governments can access hyperscalers’ infrastructure, assets, and data under certain circumstances. I suppose there’s a whole other topic there, but at least we should recognize that there's some added risk when it comes to governments and their access to this data.
Burris: It’s a double-edged sword that US companies may be worried about hyperscalers elsewhere, but companies that aren't necessarily located in the US may be concerned about using those hyperscalers because of the relationship between those hyperscalers and the US government.
These concerns have been suppressed in the grand regime of decision-making in a lot of businesses, but that doesn’t mean that it’s not a low-intensity concern that could bubble up, and perhaps, it’s one of the reasons why Alibaba is growing so fast right now.
All hyperscalers are going to have to be able to demonstrate that they can protect their clients, their customers’ data, utilizing the regime that is in place wherever the business is being operated.
All hyperscalers are going to have to be able to demonstrate that they can, in fact, protect their clients, their customers’ data, utilizing the regime that is in place wherever the business is being operated. [The rationale] for basing your business in these types of services is really immature. We have made enormous progress, but there’s a long way yet to go here, and that’s something that businesses must factor as they make decisions about how they want to incorporate a cloud strategy.
Gardner: It’s difficult enough given the variables and complexity of deciding a hybrid cloud strategy when you’re only factoring the technical issues. But, of course, now there are legal issues around data sovereignty, privacy, and intellectual property concerns. It’s complex, and it’s something that an IT organization, on its own, cannot juggle. This is something that cuts across all the different parts of a global enterprise -- their legal, marketing, security, risk avoidance and governance units -- right up to the board of directors. It’s not just a willy-nilly decision to get out a credit card and start doing cloud computing on any sustainable basis.
Burris: Well, you’re right, and too frequently it is a willy-nilly decision where a developer or a business person says, “Oh, no sweat, I am just going to grab some resources and start building something in the cloud.”
I can remember back in the mid-1990s when I would go into large media companies to meet with IT people to talk about the web, and what it would mean technically to build applications on the web. I would encounter 30 people, and five of them would be in IT and 25 of them would be in legal. They were very concerned about what it meant to put intellectual property in a digital format up on the web, because of how it could be misappropriated or how it could lose value. So, that class of concern -- or that type of concern -- is minuscule relative to the broader questions of cloud computing, of the grabbing of your data and holding it a hostage, for example.
There are a lot of considerations that are not within the traditional purview of IT, but CIOs need to start thinking about them on their own and in conjunction with their peers within the business.
Learn More About
Solutions From HPE
Gardner: We’ve certainly underlined a lot of the challenges. What about solutions? What can organizations do to prevent going too far down an alley that’s dark and misunderstood, and therefore have a difficult time adjusting?
How do we better rationalize for cloud computing decisions? Do we need better management? Do we need better visibility into what our organizations are doing or not doing? How do we architect with foresight into the larger picture, the strategic situation? What do we need to start thinking about in terms of the solutions side of some of these issues?
Cloud to business, not business to cloud
Burris: That’s a huge question, Dana. I can go on for the next six hours, but let’s start here. The first thing we tell senior executives is, don’t think about bringing your business to the cloud -- think about bringing the cloud to your business. That’s the most important thing. A lot of companies start by saying, “Oh, I want to get rid of IT, I want to move my business to the cloud.”
It’s like many of the mistakes that were made in the 1990s regarding outsourcing. When I would go back and do research on outsourcing, I discovered that a lot of the outsourcing was not driven by business needs, but driven by executive compensation schemes, literally. So, where executives were told that they would be paid on the basis of return in net assets, there was a high likelihood that the business was going to go to outsourcers to get rid of the assets, so the executives could pay themselves an enormous amount of money.
Think about how to bring the cloud to your business, and to better manage your data assets, and don't automatically default to the notion that you're going to take your business to the cloud.
The same type of thinking pertains here -- the goal is not to get rid of IT assets since those assets, generally speaking, are becoming less important features of the overall proposition of digital businesses.
Think instead about how to bring the cloud to your business, and to better manage your data assets, and don’t automatically default to the notion that you’re going to take your business to the cloud.
Every decision-maker needs to ask himself or herself, “How can I get the cloud experience wherever the data demands?” The goal of the cloud experience, which is a very, very powerful concept, ultimately needs to be able to get access to a very rich set of services associated with automation. We need visible pricing and metering, self-sufficiency, and self-service. These are all the experiences that we want out of cloud.
What we want, however, are those experiences wherever the data requires it, and that’s what’s driving hybrid cloud. We call it “true private cloud,” and the idea is of having a technology stack that provides a consistent cloud experience wherever the data has to run -- whether that’s because of IoT or because of privacy issues or because of intellectual property concerns. True private cloud is our concept for describing how the cloud experience is going to be enacted where the data requires, so that you don’t just have to move the data to get to the cloud experience.
Weaving IT all together
The third thing to note here is that ultimately this is going to lead to the most complex integration regime we’ve ever envisioned for IT. By that I mean, we are going to have applications that span Software-as-a-Service (SaaS), public cloud, IaaS services, true private cloud, legacy applications, and many other types of services that we haven’t even conceived of right now.
And understanding how to weave all of those different data sources, and all those different service sources, into coherent application framework that runs reliably and providers a continuous ongoing service to the business is essential. It must involve a degree of distribution that completely breaks most models. We’re thinking about infrastructure, architecture, but also, data management, system management, security management, and as I said earlier, all the way out to even contractual management, and vendor management.
The arrangement of resources for the classes of applications that we are going to be building in the future are going to require deep, deep, deep thinking.
That leads to the fourth thing, and that is defining the metric we’re going to use increasingly from a cost standpoint. And it is time. As the costs of computing and bandwidth continue to drop -- and they will continue to drop -- it means ultimately that the fundamental cost determinant will be, How long does it take an application to complete? How long does it take this transaction to complete? And that’s not so much a throughput question, as it is a question of, “I have all these multiple sources that each on their own are contributing some degree of time to how this piece of work finishes, and can I do that piece of work in less time if I bring some of the work, for example, in-house, and run it close to the event?”
This relationship between increasing distribution of work, increasing distribution of data, and the role that time is going to play when we think about the event that we need to manage is going to become a significant architectural concern.
The fifth issue, that really places an enormous strain on IT is how we think about backing up and restoring data. Backup/restore has been an afterthought for most of the history of the computing industry.
As we start to build these more complex applications that have more complex data sources and more complex services -- and as these applications increasingly are the basis for the business and the end-value that we’re creating -- we are not thinking about backing up devices or infrastructure or even subsystems.
We are thinking about what does it mean to backup, even more importantly, applications and even businesses. The issue becomes associated more with restoring. How do we restore applications in business across this incredibly complex arrangement of services and data locations and sources?
There's a new data regime that's emerging to support application development. How's that going to work -- the role the data scientists and analytics are going to play in working with application developers?
I listed five areas that are going to be very important. We haven’t even talked about the new regime that’s emerging to support application development and how that’s going to work. The role the data scientists and analytics are going to play in working with application developers – again, we could go on and on and on. There is a wide array of considerations, but I think all of them are going to come back to the five that I mentioned.
Gardner: That’s an excellent overview. One of the common themes that I keep hearing from you, Peter, is that there is a great unknown about the degree of complexity, the degree of risk, and a lack of maturity. We really are venturing into unknown territory in creating applications that draw on these resources, assets and data from these different clouds and deployment models.
When you have that degree of unknowns, that lack of maturity, there is a huge opportunity for a party to come in to bring in new types of management with maturity and with visibility. Who are some of the players that might fill that role? One that I am familiar with, and I think I have seen them on theCUBE is Hewlett Packard Enterprise (HPE) with what they call Project New Hybrid IT Stack. We still don’t know too much about it. I have also talked about Cloud28+, which is an ecosystem of global cloud environments that helps mitigate some of the concerns about a single hyperscaler or a handful of hyperscale providers. What’s the opportunity for a business to come in to this problem set and start to solve it? What do you think from what you’ve heard so far about Project New Hybrid IT Stack at HPE?
Key cloud players
Burris: That’s a great question, and I’m going to answer it in three parts. Part number one is, if we look back historically at the emergence of TCP/IP, TCP/IP killed the mini-computers. A lot of people like to claim it was microprocessors, and there is an element of truth to that, but many computer companies had their own proprietary networks. When companies wanted to put those networks together to build more distributed applications, the mini-computer companies said, “Yeah, just bridge our network.” That was an unsatisfyingly bad answer for the users. So along came Cisco, TCP/IP, and they flattened out all those mini-computer networks, and in the process flattened the mini-computer companies.
HPE was one of the few survivors because they embraced TCP/IP much earlier than anybody else.
We are going to need the infrastructure itself to use deep learning, machine learning, and advanced technology for determining how the infrastructure is managed, optimized, and economized.
The second thing is that to build the next generations of more complex applications -- and especially applications that involve capabilities like deep learning or machine learning with increased automation -- we are going to need the infrastructure itself to use deep learning, machine learning, and advanced technology for determining how the infrastructure is managed, optimized, and economized. That is an absolute requirement. We are not going to make progress by adding new levels of complexity and building increasingly rich applications if we don’t take full advantage of the technologies that we want to use in the applications -- inside how we run our infrastructures and run our subsystems, and do all the things we need to do from a hybrid cloud standpoint.
Ultimately, the companies are going to step up and start to flatten out some of these cloud options that are emerging. We will need companies that have significant experience with infrastructure, that really understand the problem. They need a lot of experience with a lot of different environments, not just one operating system or one cloud platform. They will need a lot of experience with these advanced applications, and have both the brainpower and the inclination to appropriately invest in those capabilities so they can build the type of platforms that we are talking about. There are not a lot of companies out there that can.
There are few out there, and certainly HPE with its New Stack initiative is one of them, and we at Wikibon are especially excited about it. It’s new, it’s immature, but HPE has a lot of piece parts that will be required to make a go of this technology. It’s going to be one of the most exciting areas of invention over the next few years. We really look forward to working with our user clients to introduce some of these technologies and innovate with them. It’s crucial to solve the next generation of problems that the world faces; we can’t move forward without some of these new classes of hybrid technologies that weave together fabrics that are capable of running any number of different application forms.
Listen to the podcast. Find it on iTunes. Get the mobile app. Read a full transcript or download a copy. Sponsor: Hewlett Packard Enterprise.
You may also be interested in: