Software Defined Networking has taken the infrastructure world by storm, and its most public face has been SD-WAN. But beyond merely replacing legacy MPLS networks, SD-WAN is being called upon to handle new challenges from technologies like IoT. With us today to talk about this intersection of new technologies is Martin Bosshardt, Founding Partner of Open Systems. Open Systems is an SD-WAN provider based in Zurich and Silicon Valley that focuses on delivering end-to-end, fully integrated SD-WAN to the global enterprise market.
TR: What does Open Systems focus on, and how did it get to where it is today?
MB: Open Systems is an SD-WAN provider that started in the financial services industry in Switzerland as a managed security service provider. Unlike our competitors, we have always operated ‘as-a-service’ and we never sold hardware and software. This really gave us the opportunity to scale SD-WAN from operational point of view, but it means we are less known in the US market and as a technology provider. We operate SD-WAN in more than 100 countries connecting approximately 2.5 million end-users, and like most SD-WAN providers we are growing relatively quickly. But we don’t just have a technology stack in SD-wan, we operate an entire platform, integrating SD-WAN and also the enterprise security stack needed it securely.
TR: What components of your platform are your own, and what components are integrated from other technology providers?
MB: The platform we operate is home grown, that’s our intellectual property. We currently integrate about 45 different products and technologies onto that platform. The SD-WAN product itself is our own stack. We have couple of routing efficiency products, like Procera, which is a very component for us to understand what is routed, why and where. Our solution is therefore a combination of different products, and that then drives the entire stack needed to upgrade a global network. That's a huge differentiator from most of system integrator competitors, who buy different products and then integrate them customer by customer. We integrate those products on one platform and deliver it always in the exact same manner to all our customers, so it's really an as-a-service approach.
TR: What types of customers do you focus on?
MB: Our sweet spot is where we can help global enterprise organizations with $2B revenue and a couple thousand people and upwards to reduce complexity on the operational side. We originally started in financial services, and about 10 years ago we entered the industrial manufacturing scene, especially chemical and pharmaceutical manufacturing. For global companies that embrace more than three or four nations, things get complicated from both a regulatory point of view and from an underlying ISP perspective. That’s really where we can add most of the value, because operating an international network gets complicated at the transport layer and different regulatory regimes add operational complexity. For example, data protection in Germany, the US, and China are all very different.
TR: What do you see as the biggest opportunity ahead right now?
MB: We see the IoT market really driving SD-WAN because it has really become a key driver for most of our customers efforts to innovate next to the cloud. It lets them connect sensors in the field and get much closer to their customers, and then use this technology to be faster and more responsive to what's happening in the market. But at the same time, it brings tremendous risks because IoTs are relatively cheap and their security implementation are relatively weak due to cost and speed-to-market.
TR: What is it about cloud and IoT that are driving deployments of SD-WAN?
MB: Until about 2012, most innovation was driven by computing power on site and Moore’s Law, but then it slowed down a bit. Right now, in the Cloud we see the computing power per dollar, is doubling about every 5-6 months. It’s really faster than Moore’s Law used to be, and this is a tremendous driver. For example Office 365 is an amazing driver for SD-WAN, because the classic MPLS networks just usually collapse. You really have to rethink the network's topology to provide a service level that meets its needs. Today we see a lot of innovation coming from machine learning, cloud-based analytics and data processing. Then obviously, if you want to use that cloud computing power for analytics, IoT becomes a very important data source in order to understand in more detail what's happening in the field. For example, machine manufacturing companies want to understand how their machines are performing in the field. IoT technology can collect this data so they can ultimately analyze it in the Cloud.
TR: What’s important in a network designed for IoT? Bandwidth? Latency?
MB: IoT usually is not necessarily a bandwidth or latency issue too much. It's more crucial that often that once you have the data in the Cloud, you have to analyze that data and do something with it. That’s why what we see very often is a combination of Cloud and IoT. Therefore, obviously, it is important also to connect IoT devices relatively cheaply to the Cloud. You really don't want to have too much effort on the network side, and therefore it really performs best if it's software-defined and with little operational effort. But this leads to the challenge we see that many companies start to connect IoT devices in an almost unmanaged way directly to the rest of their infrastructure. That can result in quite dangerous situations if you don't introduce proper zoning. We recently saw an interesting breakout a couple of weeks ago of WannaCry. It’s a very well-known virus with available patches. But the outbreak became very loud at this company because of a camera network. The company had connected to their standard SD-WAN a couple hundred cameras with an unpatched Windows stack, and one of those cameras got infected by a maintenance PC. We always suggest to use SD-WAN but introduce an IoT zone. Then you can connect IoT devices to such a zone while keeping control over what exactly is exactly communicating with what.
TR: How are enterprises approaching this right now? Are they approaching it with a plan or is it more haphazard?
MB: I think generally the latter, but it depends a lot on the company. Financial services companies are a little more cautious, probably by the nature of their business. But I would say that, in general, all IoT departments are really under pressure to drive innovation super-fast. That often doesn't give too much room to rethink the entire security architecture. We see very often that networks are migrated ‘as they are’ into the SD-WAN space. They use SD-WAN technology and profit from all the capabilities, the technical capability and the flexibility. The rethinking of the security approach to that is needed is usually only done later when the first incidents happen. IoT really can be dangerous if it's not managed properly and can create a lot of operational effort.
TR: Other than security, what types of problems can arise for enterprises having put IoT into their networks in some form?
MB: It is relatively hard to debug. First of all, in an SD-WAN you cannot use protocol-based routing anymore. It is application-based routing and that adds complexity. Then, an IoT infrastructure can perform very well in one area of the network, but in another area doesn't perform well. To really understand what is causing the performance issue takes an entirely new set of tools. This is really where we see many do-it-yourself driven companies, which is still the majority of large enterprise organizations, have trouble. It is relatively hard for them then to come up with the tools to debug the performance issues.
TR: How does Open Systems’ solution help enterprises address problems of this sort?
MB: One main advantage we see is that because we integrate the security protocols and the SD-WAN part, it is easier to follow a packet. If you track a packet from A to B through five or six different products, each with a different graphical user interface, it's really hard to understand what exactly is happening. We can provide a lot of value by integrating them, giving much faster information about what happening to those packets in each step of their path around the world. This is mainly because we are in not as technology driven. We come from the operational side, and we really take operational responsibility.
TR: What can enterprises do today to run better networks for IoT?
MB: I think that zoning is the most important part at this point. Most companies don't have proper zoning, and this is an aspect of security that is important in the physical world too. The Titanic is a very good example of where zoning was introduced but not implemented in the proper way. Modern ships don't sink like that anymore because the zoning is done properly; those chambers have been pressurized. The same is true when it comes to fire safety, and is now becoming true for core IT.
TR: Where do you see technology evolving further in enabling enterprises to deal with such problems?
MB: We have to not just keep the enemy out but also really understand what's happening inside the network. One area that technology still needs to improve a lot is to identify anomalies much faster than we are able to do today. There is such functionality in place, but I see great potential there to really improve early alerts and detect malicious behavior of network components much faster.
TR: What role do you think artificial intelligence and machine learning will play in this?
MB: I think AI could help a lot. If you say 80- 90% of traffic is not malicious, AI can help to identify malicious traffic of unknown patterns much faster. Right now, most malicious traffic is detected by feeds and known patterns. Everything that is unknown is not recognized fast enough and machine learning can help with that.
TR: Will you be looking to solve this problem yourselves, or within a broader industry framework?
MB: Both. This entire space is huge, and it would be strange to believe that we can solve all the problems ourselves. Where we have a certain strategic advantage is that Open Systems is operating, we believe, the largest fully standardized SD-WAN platform in the world. Most other products, even when very large, are not operating in a standardized way. 80% of our incidents are handled by robots, while just 20% are escalated to engineers and operators. This very high level of automation and standardization makes it easier to apply AI to find malicious traffic. In some cases, it already works very well. But right now, we really have to design those cases carefully and then run the algorithm against those scenarios, after which we find those scenarios very accurately. But today’s machine learning algorithms won’t find patterns we haven't thought of before. That's really the innovation we need to move.
TR: AI is one of today’s biggest buzz phrases, but how far do you think we are from living up to the term?
MB: Artificial intelligence is a dangerous phrase to use because I think it's not really intelligence today. AI today is brute force correlation, and it’s truly fascinating what you can do with brute force correlation. When I was studying electrical engineering, I worked with neural networks to recognize snow and rain from satellite pictures, and it was fascinating that it was possible but it’s not intelligent. Whenever you see a self-driving car, it's fantastic tech but it's not magic. The cloud is really helping to drive these brute force correlation engines on a crazy level. It is powerful, absolutely, for example in medical research where you can sequence DNA within 24 hours and then correlate that DNA sequence against the cancer DNA and find the differences. This is phenomenal and it can surely help find malicious traffic in networks, but it’s not intelligent in the way we think of it yet, unfortunately.
TR: Thank you for talking with Telecom Ramblings!
If you haven't already, please take our Reader Survey! Just 3 questions to help us better understand who is reading Telecom Ramblings so we can serve you better!Categories: Artificial Intelligence · Industry Spotlight · IoT, M2M · SDN