Podcast #8: Scaling Analytics for the IoT era

Hello everyone, and welcome back to another episode of the Momenta Edge Podcast, this is Ed Maguire, Insights Partner. Today we have Kate Mitchell who co-founder and CEO of Edge Intelligence joining us. 

Kate has a really interesting background in technology as an investor, as an executive, as a leader in innovation, and we’d like to talk about what you’re doing now, and then go into some of the context that has brought you to focus on Edge Intelligence. So, Kate tell us a little bit about what Edge Intelligence is doing. 

Edge Intelligence is actually an evolution of an idea that we came up with about 9 years ago, and that was around the core of how you get analytics in the way that is useful for any business executive, any user to make it much timelier, and much more accessible. It comes from the 1970s where everything was centralized, and architectures were somewhat rigid, and reporting at that time for instance was monthly. The internet hadn’t been invented, and so many things have changed over time that the underlying architecture’s data, roll architectures for databases, and column architectures really don’t solve the problem given the tremendous line of data, and the marketisation, and the decentralisation of data. 

So, we came up with the idea that there had to be a better way to do analytics, and we started out focusing on the database component, and what’s the fundamental way that you store and access data that would make it much more accessible and would handle and easily scale to the lines that we see today. So that’s where it all began. 

That’s great, that really hits in many respects on the Holy Grail of Connected Industry, trying to extract value from the data that gets generated by all these billions of devices that are connected, and complex systems, and be able to get the right information to the right people, at the right time.  

You have a really interesting background, and when we spoke before we initially connected around our Edge Intelligence, or Intelligent Edge rather, webinar. It was quite an appropriate mix of topics and names. When we spoke, you told me a bit about your background, and I thought it was really fascination to hear a bit about what you’ve done in the past, and how that’s brought you to focus on the problems that you’re addressing with Edge Intelligence. Tell us a little bit about your background. 

It started at IBM, typical sales and marketing branch manager, working for the Chairman of IBM, running an industry organisation where the whole goal was to figure out how do you handle information, and of course this was in the earlier years, going back to the late seventies, and the 1980s, and figuring out how do customers across a broad range of industries put data to work? How do they access that? How do they get insight from it? How do they take action? Then after about 14 years there, a great-great time at IBM, I decided it was time to go do something a little more entrepreneurial and went out to California, worked again as one of the first companies focused on analytics, at Metaphor Computer Systems. At that time there were no PCs that were appropriate/affordable, so the company when I joined them, they were making hardware as well as the analytic software, working closely with IMB at the time, and in fact ultimately IBM bought that company. 

Then I said, ‘I really like being out on the West Coast, let me see what else is here’, and I ended up getting hired at Oracle. I was hired to run marketing there, I did that for a number of years, so really became immersed and focused on data. So, this theme of how do you manage major implementations, and then how do you start to think what are the applications that need to be written? What’s the logic of surrounding the capture of data, and providing ultimately not just the systems that are the transaction operational system, which of course are a key requirement, but after you’ve captured all that data, how do you make sense of that? 

The whole idea of a separate analytics system and data warehouses certainly became the main theme, and still with us today, decades later. Many people still get their information from what’s become a somewhat rigid data warehouse, and I think our view at Edge Intelligence is, it’s really time for this democratization of data, and make it useful to anybody at any time, and very accessible, very agile too. Because when you’re starting something, you don’t know what the data is going to look like five years from now, you don’t know what the queries are going to look like. So, ideally you want a system that hides all of that complexity and takes care of all of that for you, so you can focus on running the business and improving your business. 

You were at Oracle at a pretty pivotal time in the database market, and anybody who has studied the competitive dynamics of technology, for instance looking at Geoffrey Moore’s work on ‘Crossing the Chasm’ and ‘The Gorilla Game’ for instance; that was a seminal time in the development of the software industry, and you had an up-close view of what was going on in the database market, which has become really that classic example of where you had one company, Oracle, that ends up dominating and becoming the gorilla in the market. What were some of the lessons you learned, or any parallels from the competitive landscape in that time which are relevant today. Were there any actions that Oracle took whilst you were there which really stand out in your mind, that helped them prevail over all the competition, Informix, Red Bricks and Sybase’s of the world that were all vying for a piece of that growing pie? 

It’s funny you mention Geoffrey Moore because this was before he had written the book, ‘Crossing the Chasm’, but he wasn’t quite the household name that he has become, and I was able to have him lead one of my planning meetings. I had executives from around the world that ran marketing in various countries and regions, and Geoffrey came in to run the session for us. At the time it was such a competitive space, we spent a lot of our time, and I spent a lot of my advertising dollars, going after Sybase, it was all about performance at the time. Of course, Larry always loved performance, he was the consummate marketer, in some ways it was a lot of fun because here is the guy who said, ‘I hate marketing, keep it away from me’, and yet he wanted to be in any meeting that was strategic where we were talking about direction. He always owned the product strategy, there was no doubt in anybody’s mind he owned the product strategy.  

But in terms of the marketing dollars and where were we going to focus our attention, and who were the real competitors, as I’ve said, at the time it was all about performance, it was all about leapfrogging from a technology standpoint, and Sybase was clearly in the sites. But Informix at the time, hard to believe it now, Informix had some pretty amazing software, so from the performance standpoint we always thought that Informix might turn out to be one of the companies that was a major competitor.  

A couple of things I learned there was, there’s no substitution for just that passion about what you’re doing. Even at the time Larry reading resumes of people that were about to be made job offers, even relatively low entry level employees, he just felt so passionate about this company that he had built. It all revolved around, ‘Are you going to be extremely confident and carry out the vision that we have? Whether you’re developing the product, supporting the product, selling the product, or marketing the product, just really focus on ‘I’m gonna build this company into the leader and the world-class organisation’. Also, couple with that, focus on the competition. 

I think it’s what drives every company today is, ‘Who are my competitors?’ and ‘How can I outsmart them?’ And in so many places, interestingly, obviously for a database software company that’s a core component that you’re looking to make better than the other guy. But for every organisation I think of, or even government public service entities, public sector entities, it’s all about, ‘How do I understand my business, so I can improve it?’ Whether that’s, how do I get more creative, so I get a bigger share of the pie from a revenue perspective, or how do I get more creative, I get classed out? I’m really smart there and I become the low-cost leader, so my profits are growing nicely, or how do I provide the best customer service that’s out there so I can avoid churn? Mobile providers, people will like my services, they’ll like the quality of the service, they’ll like the pricing, they’ll like the packaging and what they can buy, and that maniacal focus on what do I do to make my business better, to me it was a great lesson, it all comes down to data, how do I use data? 

It’s a great point you make there, that it ultimately ties into customer satisfaction. I think what’s notable, you had just called out how Larry was very much focused on performance as an aspect of embarking, and I think as I’ve listened to him over the years as well, I think he naturally gravitates toward touting the performance of a database; because when your customers are IT buyers and database administrators, they obviously are going to want to get the best performance for their dollars, and it becomes very much of a tactical sale. But as you alluded to right at the beginning of our conversation, analytics is about business value, and getting to the business user, and that’s a different type of sale.  

Going back to the genesis of data warehousing, executive information systems, and then the emergence of business intelligence, how did you see that tension play out the focus on getting the technical aspects of a solution right, with the database, appealing to the IT people; but then how did that messaging to business people play out effectively? What kind of challenges did you see in the market as this has evolved? 

In the beginning of course, the thing about the database it was companies that were built, and that’s what they did, they offered databases. Then we and the competitors realised there are companies out there that are offering applications, and now they need the database underneath that to drive the application. I remember at the time, SAP was a great example, because they were getting a lot of their revenue from the applications; we said at the time, ‘We need to be in the applications business’, and of course that was the beginning of a whole new wave of investment, and a product suite portfolio that we could offer, that said it’s not just about the database. The database of course is the enabling technology, it’s the core component of the infrastructure, but it really comes down to what’s the application that will drive the database, and will deliver the results that are needed?  

So, suddenly you saw some companies moving from being just purely database companies, to saying, ‘I’m going to add the applications on top’. But it really comes back to the transaction, the operational system on one side, and capturing all that data, and it powers every company and government entity today. Then there’s the other side which is the analytics side, that’s always viewed as secondary, although I think over time it’s been viewed as much more important to get the insights from that data and be able to do it in a way that is easy to do, it can be done in real-time. I think the who concept of the Internet of Things is changing that dramatically as well because suddenly you’re dealing with 20, 30, 40 billion IoT devices being connected, 600 zettabytes of data from IoT being generated every year, and you just can’t say, ‘Alright, business as usual, we’re gonna move all this data that’s been generated to a highly-centralized database, and then we’re gonna perform analytics, getting all that data from where it begins life out at the edge, at the factory floor, or the hospital, or the aircraft carrier, we’re gonna bring all that data back to either the central datacenter, or to a cloud, that just doesn’t make sense. 

So, that’s really what we’re looking at now is, how do you make that tremendous jump from redefining data that you normally deal with in one of your multiple large corporate data centers, enhancing that with data that you keep in the cloud. The next step is, what about the data that’s at the perimeter of your organisation, maybe it’s even a virtual perimeter, but it’s out there; it’s in every remote location, every retail store, every branch office of a bank, every factory, how do I treat that as if I still have control, and it’s all physically resident within my corporate data center, or in the cloud? 

That’s really where the challenge comes in, and hence our evolution as a company from thinking about how do you just do analytics better where you store and access data in a very agile way, with highly-highly scalable, where we’ve extended that in the last 18-months to say, how do you think about networking all that data, and securing it, whether it’s at rest or in-flight? So that data can reside anywhere on the network and without physically having to move it, you can reach out to do analytics to it, or to do distributed machine learning for instance, leading it right where it begins life, and getting away from all of the concerns about latency, cost of moving it, privacy, or geopolitical concerns that come into play when you’re talking about moving that data from its origination point. 

What’s been so interesting as we fast-forward through a couple of decades after the original database wars, is we saw the applications turn into these stacks, and over time you had more and more analytics that would get embedded into a stack where you’d have Microsoft and SAP, and IBM to a lesser extent, although they were not as heavy on the application side, and Oracle; basically arguing that the way to go was to have analytics that were embedded in your applications, and had that integrated stack. But then as you move into more of a cloud model, customers don’t want to be beholden to that single vendor to the same extent. You’ve got essentially a centralisation of architecture with this cloud model, and of course the decline in cost of sensors and processing, and edge devices, the ability to instrument these physical processes, does create a far more distributed environment. 

So, now as we’ve gone to a cloud environment, you had started Edge Intelligence a few years back, but when we think to around 2006/07, the rise of BigTable and Hadoop emerged around 2009, we’ve seen this whole emergence of a lot of different types of databases, which in a sense we had this consolidation around a few of these relational databases, and then all of a sudden there was this explosion of innovation. I’d love to get your thoughts on what drove the market to evolve like that, and how you see that parallel with the explosion of software as a service, and cloud, and the types of challenges that’s presented to the companies in the market, trying to figure out the right solution for the right problem. 

I think the rise of the NoSQL which some people think means no sequel, it means not only sequel. I think the rise of that was because of the explosion of volume of data, relational databases simply weren’t appropriate for that. You didn’t need all that referential integrity, and two-phase commit to make sure that every single thing was precisely as it should be, otherwise it got rolled-back. Which you need to have in financial transactions of course, you don’t need that when a lot of data is social media and so-forth. So, I think the idea of the NoSQL databases, including Hadoop emerged with the idea that there’s a whole different type of data, and not only that, the purposes that it’s being used for are completely different. 

So, we’ve got to come up with alternate views of how we’re gonna create databases, how we’re going to store it, access it, process it. And I think at least for businesses and many big government entities, what we started to see is even with something like Impala on Cloudera, the realization that, ‘Wow, I’ve got to be able to handle that structure query language, there’s a demand for that’. So, our view all along is, look the important thing is, how do you ingest data at network speed? If you just think about what are the requirements for analytics today? You’ve got to be able to ingest the data at network speed, process the complex events in near real-time and process them usually at the Edge. It needs to be cost-effective for storage whether you’re storing gigabytes, kilobytes, or petabytes of data, just maybe thinking about small physical footprint. Some use cases you need to retain the data from a few minutes or hours, to days, weeks, months, and the cloud model does not really support that very well, it starts to get very expensive when you talk about keeping a lot of data for a long period of time. 

On the other hand, there’s some analytics applications where the important patterns that you’re looking for don’t become evident with a relatively small slice of the data, which is only retained for a short period of time. So, we’re starting to see some of those requirements where even what some people would think today is pretty advanced, some of the analytics you can do in the cloud; actually, it’s good for certain things but, there’s a whole category of use cases and applications that come back to, I don’t want a sub-set of the data, I need all of it. I know I can’t physically move it, and I’ve got to find a way to cost affordably without big IT investments, or database administrators. I’ve got to be able to handle all of this in a distributed way, whether that’s Internet of Things, whether its network monitoring, whether its predictive analytics, security, surveillance, customer supplier analytics, including being able to look at your partners, and being able to look through that supply chain and figure out where there might be bottlenecks, or where you might need to make some changes.  

So, I think the systems generally reflect the nature of the business problem you’re trying to solve, and I think what we’ve seen is this evolution over time from the central monolithic approach, to a distributed approach, to the rise of the cloud, the AWS and the Azures of the world to say, ‘I’m going to offload my corporate data, I’m going to get the data that I want and accessible in the timeframe that I need, in a format that I need to make the decisions, and people going off and doing their own thing; to now, wait a second, there are other ways to do this that are very affordable for lots of data, long-term, and understanding the idea of the central data warehouse, that concept really is dead from all vantage points. 

Yes, that’s the old Teradata model of the enterprise data warehouse, essentially where you would have this master record. Of course, having a single view of the truth from a business standpoint is critical, but this idea that you’re gonna be able to access, to have all the analytics that you need for all of these different processes that are distributed, it’s really difficult to manage that with the latency challenges, and the volumes of data that are emerging. 

So, I have a question about applications, and how you see businesses needing to rethink how they architect their applications? Again, as the pendulum swings from centralized to de-centralized, to centralized, which was we could consider that cloud mobile model to be centralized, to de-centralized again; are companies going to need to rethink and to rearchitect their applications, either their older applications, or does there need to be a new way to think about how you architect your new applications, to take into account the capabilities of being able to have some processing on the edge, and manage this entire continuum of processing and analytics, and data movement capabilities across the full continuum? 

If you look at it from a vendor point of view, there are some vendors with a legacy approach, it’s just not possible to get from where they are to where they’ll need to be in three to five years. They struggle I think trying to find ways to say, ‘Okay, now we also have a cloud approach’. But I think when it comes to being able to capture and access that data at the edge, and retain it there, so you can do the analytics on data whether it’s a second-old, or a year-old, I think that kind of approach needs to start from the ground, up. And our view is, of course we’re happy to work with application vendors, to say, the big challenge there is what is the database, and data management underneath the application itself, underneath the logic, that will help support that and drive that. That will handle all of the requirements of that very fast network ingestion, which is something that for instance Hadoop is great at, not so great at being able to slice and dice, and find exactly what you want, and get exactly what you want out of the data that’s stored. Give you the best of what would be for forensic query a road store where you would be looking for a needle in the haystack. Normally a road store would be excellent for that, but not very good at saying, ‘I want to look at all my subscribers in this postal code, or zip code with this kind of demographic information, salary and so-forth, number of children and where else they make purchases, to figure out touching all the data, to figure out what kind of special pricing and packaging shall I offer specific subscribers that are ready to make a change, so that I avoid losing them? 

So, the database is really at the heart of how do you support that application? And I think some of the existing application companies are working very hard to move their application into the current time, and looking ahead to, the source of the data is no longer a single source in a corporate enterprise, in a corporate data centre. And even thinking maybe there’s two sources, maybe it’s a network of corporate data centres and network of cloud data centres; that gets you to where a lot of customers are right now, but it doesn’t get you three to five years ahead, where some of the more innovative customers are pushing the boundaries and saying, ‘This thing has got to be able to scale, and it’s got to be able to handle data anywhere in the world that makes sense to me.’ As long as I can make it accessible in real-time, and keep it indefinitely, that’s what the future looks like, certainly for IoT and connected industry, and that’s what I’m going to demand of my application, and the underlying data management infrastructure that will support that. 

This is a great opportunity to ask you more specifically about what Edge Intelligence is doing, could you provide a little bit of context of the types of problems that your technology is specifically well suited to address, or I should say uniquely well suited to address? 

It’s interesting, it was one of our largest customers in Asia, and this is when we were just one database before we had enhanced the product and changed the name of the company. This is going back a couple of years, where the government entity was worried about individuals, and bad behaviour in individuals, they had an old-fashioned way for monitoring that. They said, ‘What we really want to do is we want to monitor individuals, we want to monitor not only their cell-phone records, but we want to monitor their physical location’. So, that data for one hundred-million plus citizens is so voluminous that we can’t actually bring that from the telecommunication providers to a central site. So, what we really want is, we want to be able to leave all that data where it begins life, at the network providers, the big telecommunication providers, leave it there and we’ll issue queries from a central site, we’ll issue it and go out across the network, and the results will only bring back the relevant information. 

This has two benefits, 1) It’s not going to burden the network and be hugely expensive, and have huge latency, but 2) It’s really going to respect the privacy of law-abiding citizens. So, we implemented that with the small systems integration partner in Asia, they’ve asked us not to talk about the name of the country because it’s a stealth implementation, but it’s the largest one we have. We looked at that, and we said, ‘You know, that starts to look like the Internet of Things, where you’ve got data in many different places, and you don’t want to have to move all that data in order to get the insight’. So, that’s when we said, ‘Let’s go build this next layer of capability on top of the product. Let’s go build the orchestration and management of all of this data that can’t be managed from a single site, but you don’t have to move the data to a central site’. 

It required some major new changes in technology in order to be able to say, there’s the processing side, we’ve got to be able to have network speed ingest, complex event processing in neuro time at the Edge. There’s a storage challenge of how you think about that, it goes from small amounts of storage, up to enormous amounts. Those are billions of transactions each day being kept for several years, that gets into petabyte size. You need a small physical footprint, you need to be autonomous, you can’t deploy technology-savvy people at the Edge in order to worry about that, or maintain that system, it’s got to be completely autonomous. And of course, the final thing is, when you go to do the analytics, you want to be able to respond very quickly to any adhoc query, you want to be able to handle very large datasets, trillions of records, and maybe the most important thing is, there’s always a changing nature as your business evolves into what data you want to use, and how you want to use it, how you want to analyse that. 

You don’t know what that is up-front, you’d like to try and plan it, but in the old-fashioned way of that data warehouse you really had to know, what data you’re using, you had to model it, then ‘These are the types of queries I can do’, and then getting fresh data into that data warehouse was always a challenge. That’s over, now you want to be able to say, ‘I’m going to deploy something, and it’s going to be very agile, it will handle any type of query, and it will handle it very-very fast, and I don’t have to go back and change anything as my data’s evolving, or I come up with new queries. What I’ve installed here is completely agile and there’s no design-work to be done by the user’. I think that’s the new standard for just make it easy to get at the data wherever it is, make it cost-effective, and the key thing is to make the analytics work for you; bring the analytics to the data, don’t take the data to the analytics. I think that’s what will really empower end-users and will allow them to find those areas where they can make changes, and that company or government entity will operate in a far better way and be innovative on the Edge and will capture market share as a result. I think that’s really the bottom line. 

Well that’s a real paradigm shift change in mindset. I’m interested to get your take on what you see as real challenges, hurdles, or obstacles for this mindset or this approach, to result in really successfully embracing all of the new technologies and the processes that are involved in realising some of the vision. Essentially bringing the data to the user; what are big hurdles that you see, whether they’re technological, organisational, or cultural? 

You always have a cultural side of things, and you always have the early adopters, people who have the mindset that says, ‘I’m gonna go out there and take a risk, and I’m gonna explore. I’m gonna find a new way to do things’, because, we just can’t do things effectively. We’ve got as much Scotch tape on this thing as we can, and we’ve bolted on as many things as we can, but it’s just not gonna handle our needs going forwards. So, it’s really those innovative leaders that have to think about what the businesses processes are. IT is not longer, ‘Oh yeah, it’s one thing we have to worry about, it’s a cost we have to deal with’. Suddenly you’ve got to have this as the lifeblood of your company, it’s got to be that realization that the CIO or your CTO doesn’t report to your CFO, it’s not a cost center to be managed!  He reports right to the CEO, and every business leader has to have a vested interest, and has to be measured on how are you going to do things differently as the needs are changing, as the competitors are changing, as technology is changing, how are you going to do business differently? What’s it gonna cost, and what are the expectations for what we can get out of that in the six-month timeframe, the one-year timeframe, the 18-month timeframe? 

Sometimes I think a lot of us just say, ‘It’s good enough’, we’ve got to this point, it’s good enough, let’s go just now take advantage of that, and let’s coast for a while on the additional revenue we’re getting, or the additional profits because of the changes we’ve made’. But those days are over, you can’t coast on that, some good decisions that were made and some nice results, you’ve got to be constantly looking ahead, and saying, ‘This job is never over, and it only gets more difficult. But I’ve got to be one of those people who is looking forward’, sometimes we say laughingly, ‘Living on the Edge!’ You’ve got to be able to sort of force yourself to look around to say, ‘Are there other things I could be doing that would give me better results’, without the downside of trying every new thing that comes along, and never really having a change to judiciously try things and make sure they work before going onto the next thing. There’s always the fine balance that needs to be struck. 

As you look forward are there any technologies that jump out at you as showing enormous promise? We did a webinar recently on combinatorial innovations around AI, Augmented Reality, and Blockchain. As you look at the evolution of analytics, business analytics, are there any innovations that you see as particularly promising? 

I think the whole focus on, obviously AI and machine learning is a component of that. I’m happy to see how many mainstream companies are saying, ‘I’m going to use blockchain, I’m going to find a way to take advantage of that’. I thought there’d be a lot more resistance to that, and so I think that shows you that at least many of the leaders there are saying, ‘There’s some risk associated with it, and its not well-defined but, I think it can redefine the way I deal with customers, or suppliers, that can redefine my business, and I’m going to embrace that’. I think there are technologies right now which are not at the point where they’re on the scary Edge, or the bleeding Edge, which can make a big difference. I think its great to see companies taking advantage of that. 

I think it comes back to, are companies on their own able to do that? Or, do they wait and take the lead from their preferred systems integrators that might have more experience, looking at a broad set of companies across their industry? I think those systems integrators, part of their life is exploring those new technologies, and figuring out which ones will really make a difference, and helping usher those into the general marketplace, and getting companies comfortable that they’re not taking such a high risk.  

I think we’re just at the beginning of seeing the innovation ahead of us, whether it comes from this idea with the concept of IoT where all these connected devices letting you monitor everything in real-time and take action. Or Artificial Intelligence, and just moving toward how do you get the ideal combination of people, machines, and resources, so that you’re basically reinventing industries, you’re reinventing companies. Its exciting to be part of that, its an exciting time to be in the business. 

It is, no doubt, and that’s what we focus on, and we’re passionate about at Momenta.  

Kate, this has been a terrific discussion, I really appreciate your taking the time, and value your insights. I think its been enormously informative, and I want to thank you for coming on. 

Thanks for the opportunity, I love working with Momenta, you guys are great. 

Terrific. Again, this is Ed Maguire, Insights Partner with Momenta, and our guest today has been Kate Mitchell, co-founder and CEO of Edge Intelligence. 

Thank you for listening, and if you have any follow-up questions, where can people find you Kate? 

They can find me either at the website, or they can find me if they want to chat quickly, shoot me an email, Kate.Mitchell@EdgeIntelligence.com. 

Anybody who wants to reach me can contact me at Edge@Momenta.Partners, we’re on Twitter, LinkedIn, and Facebook, and all-over social media. So, thank you so much for the time, and we look forward to continuing our conversations.