We sat down with Marcio Saito, CTO of Opengear to talk about the importance of networking in today’s cloud-native space and how NetOps is more critical today than ever before.
Here is the rush transcript of the interview:
Swapnil Bhartiya: Before we deep dive into the COVID-19 and what challenges businesses are facing, I want to understand a bit about Opengear, can you tell us just a bit about what does the company do?
Marcio Saito: Yes, Opengear is a company that has been in business for about 15 years and over the last several years, it became the leader in what’s called out-of-band management. Out-of-band management is a system that enables network engineers, service providers to manage their network infrastructure, even when the production network is down. This is really what we do, is a system that helps people to manage large network infrastructures.
Swapnil Bhartiya: What kind of challenges this crisis is creating, especially for the network infrastructure?
Marcio Saito: If you’re managing servers or a storage or cloud infrastructure, and the underlying network infrastructure is working, then you can work from anywhere. You can use your management tools, the management tools can reach all the devices and you’re fine. The network infrastructure is really the lifeline for everyone else in IT, because everyone relies on the network to do their job. Having a resilient network is very important, particularly in these times of crisis where the coronavirus, what it does, it basically forces everyone to work from home. If you are in a virtual environment and you have an internet connection, you can work from home and that’s not a big deal.
The difference is that in networking there is always the need to connect point A to point B. There is always a geographical, physical connection between the work you do in the logical layer and the physical layer, no matter how much you virtualize, how much you automate, network engineers, they do need to have access and connect to physical devices at remote sites and that’s what makes network infrastructure so special. If the network engineers cannot connect physically to a device, then they cannot do their work. That’s why sometimes it’s very difficult if you’re managing a network infrastructure, to be able to work from home, you need to be able to connect to physical devices, even if your internet connection or even if that site becomes disconnected from the network.
Swapnil Bhartiya: Working from home is just one aspect. Personally, I have been just ordering things over Instacart or Amazon Prime, because I just don’t want to go. A lot of businesses which we’re relying on physical or other activities, banking and everything else, everything is moving online, which also means that for those businesses to continue to offer their services seamlessly, without an interruption, number one. Number two is in the cloud world we always talk about high availability, data recovery, but the thing is that network, as you said, that is the lifeline. Without it, when your two data centers are not connected, there is no backup, there is no resiliency there.
Just from your perspective, because when we talk about cloud, we talk about everything, but we don’t talk much about networking and that is a lot. How critical is network infrastructure for this business continuance, whether it’s Amazon or everybody’s watching Netflix. These are entertainment, but there are a lot of critical infrastructure that relies on networking as well. What role does network infrastructure play so that these businesses can continue to offer their services without interruption?
Marcio Saito: I think most business that are managing a cloud infrastructure providing service to consumers and to business, they are designing their infrastructure to accommodate changes in demand. If you’re in retail, you have to be dimensioned so that you can scale your services during Black Friday week, for example. I think every business has seasonal variations in demand, and most people designing cloud infrastructures, they have thought about that and they have infrastructure that can actually scale to cover the demand.
What happens is one thing that is more difficult to scale compared to web servers and just compute nodes is networking. As I mentioned before, in networking there is always the connection between point a and point B, so there is a physical infrastructure that needs to be built. You cannot just push a button and create a new virtual network. The virtual network requires a physical infrastructure underneath and building out a physical infrastructure takes time and money. The big challenge all the service providers are facing today is that they can scale their compute nodes, but they cannot scale the networking at the same level.
When everyone is trying to access the services from home or the demand for online shopping increases all at the same time, it doesn’t follow the patterns of the demand from the past because everyone is rushing to the resource at the same time. The big challenge service providers are finding today is how to scale the network fast. We can see that in our business though, in the last several weeks, the enterprise business has gone a little bit quieter and slow, the business on the service providers is booming when it comes to networking equipment and management tools, because they are trying to add network capacity as fast as they can.
Swapnil Bhartiya: As you were explaining earlier about Opengear, how does Opengear tackle this problem?
Marcio Saito: As I said before, the network is really the lifeline for everyone else in IT, but if you are a network engineer and you’re responsible for maintaining the network running, when something goes wrong and you lose the connection to the network or a specific site goes offline, Opengear creates a system that is the lifeline for the network management people. What we do is we create a system that is separate from your production network, it relies on LTE cellular connectivity and basically the developer position of our system is I’m a network engineer, I go to a Opengear and I say, “I need to reach the management part of my router number three, in the London data center.”
I don’t need to know what the state of the network is, I don’t need to know if the internet is up or down, I don’t need to know if the data center is up or down. I go to Opengear and Opengear takes me there. That’s the value proposition for our system today and that’s how we help network engineers to manage their own network.
Swapnil Bhartiya: I want to understand, you explained what the company does, but I want to understand the core components of Opengear. How much of it is hardware, how much of it is software?
Marcio Saito: One principle we have to scale and to make a network more resilient, particularly if you’re talking about Wide Area Networking where you cannot create redundancy for every single component of your infrastructure. If you’re in a data center, the traditional way to increase availability is redundancy. You just add more, if are concerned about one switch failing, you put two switches in parallel. If you are working a geographically distributed network, redundancy sometimes is very expensive and sometimes it’s not even viable.
What you need to do is to take a different approach, to follow a different strategy to be able to manage a distributed network. Actually, we need to borrow from the telco industry, the telco industry has been managing distributed networks for decades and the secret for building a resilient network infrastructure is really to separate the data plane from the control plane, as well as the management plane. If you have those infrastructures separate and independent, then if there is a failure in the data plane, because there was a security attack, or because there was a connectivity problem, your management plane is still working, and it still enables you to monitor, to manage and to remediate problems out-of-band.
The concept of out-of-band access, out-of-band management came from the telco industry where you have the phone lines that provides the voice connectivity, but the switches they talk to on a separate infrastructure using separate protocols to connect the calls and to manage and monitor the quality of the calls. When networking goes to the edge, you need to take a different approach than you take in a data center by separating the management plane from the data and control planes. That’s really the strategy we follow, we recommend, and we implement to our customers.
Swapnil Bhartiya: We have talked mostly about coronavirus here, but in general, when it comes to network resilience, there are a lot of challenges. You’re in the best position to talk about that, can you talk about what are some of the top challenges that companies face when they try to deliver end-to-end network resiliency?
Marcio Saito: There is an example I like to give, is an IoT example. If you think about IoT today, the questions people ask to an IoT system are things like, “I want to monitor the temperature at that site, so every five minutes just collect the temperature and send to me.” If you are doing that, you can do that from a cloud server and basically every five minutes, there is a data packet coming from the edge to the cloud infrastructure. Now, what’s going to happen in the future is that the questions we are going to ask our systems is going to be different.
Imagine you have cameras deployed over the all traffic lights in a city. Let’s say, you want to ask you the question, “Camera, I want you to tell me when a police car crosses this intersection, for example.” One way of doing it is you have the video camera streaming video, 24 by seven to the central site and you would have tens of thousands of cameras doing that. That of course consumes a lot of bandwidth and is just not practical. What’s going to happen is you’re going to have to push the algorithms, the intelligence to detect patterns to the very edge of the network. This is what’s happening in IoT, this is what’s happening with a set of other next generation applications, where you need more intelligence at the edge.
The challenge you have there is you have to deploy more compute resources, close to the end user, farther from you, farther from the network operation center. You have to rely on wider networking connections, which are not as resilient as a network connection within the data center. So you’re facing the problem of less reliable connections and you’re facing the problem of having to deploy computer nodes at the edge of the network where you cannot have a lot of freedom.
If you have a small data center at the top of a cellular tower for example, you might have 20 servers there, you cannot just say, “Well, if a server fails, I’ll just move the workload to the other one.” If you have tens of thousands of servers, that’s a valid proposition, but not if you have 20 servers on top of a cell tower. That’s really the big challenge people have when they are managing network infrastructure and the workloads are moving to the edge.
Swapnil Bhartiya: When you look at computing, as you also mentioned that a lot of decisions are being made at the edge for a lot of reasons, it could be the privacy, it could be latency, it could be, “Why are you moving all the data from here to there?” At the same time, edge is also becoming a kind of data center in a way. What unique either opportunities or challenges edge poses for networking or for a company like Opengear?
Marcio Saito: When we talk about the edge, we are talking about the devices very close to the end user, so all the way from their mobile phones to IT equipment inside branch offices and things like that. For the service providers, for the companies that are providing cloud services to the consumer, and let’s take video streaming as an example. We know today that if companies like Netflix were trying to stream video all the way from a central location, they would consume all the bandwidth available on the internet. What they need to do, or what Google and Microsoft and Nintendo need to do, if they want to do game streaming is they need to build these data centers that are edge data centers. They’re very close to the end user, they might be a small data center in your neighborhood, it might be data center that sat on top of a cell tower, it might be a data center that sat inside the traditional telco central office.
What we are going to see in the next decade is the multiplication of those edge data centers, which require very different approaches in managing, in maintaining, because in a data center, if something physical breaks, you just send an engineer there and they can replace the equipment. If you have servers at the top of a cell tower, they might be in very remote locations where you just cannot afford to roll a truck and send someone there. The approaches you need to take, they are different, it’s less about redundancy, is more about separation of planes. It’s about being able to remotely connect and manage even at physical level, without having the need to send someone there.
That creates a lot of challenges that the industry is coping with. We are learning how to manage edge infrastructure and creating a lot of opportunities for companies like Opengear, who are creating the systems that enable that. With Opengear today you can connect network engineers to devices at the edge, so people to parts. But what we are developing and what the vision for our system is, is to evolve out-of-band management systems from being the connection between people and parts, to being the system that extends the reach of automation and management systems, so that they can reach the edge even if the production network is down.
That’s really the next step in out-of-band management, is how do I help configuration management systems like a Cisco DNA center, for example, or log management systems like Splunk or traditional monitoring systems like SolarWinds? All those tools, now they need to manage infrastructure at the edge and if a remote site goes down, that’s exactly when you need to manage that site and the traditional approach doesn’t work because suddenly if the network is not there, your management doesn’t work either. Opengear is expanding the scope of the out-of-band management system, not only to connect people to parts, but to extend the reach of those systems to the edge of the network.
Swapnil Bhartiya: Talk a bit about the trends that you’re seeing in terms of NetOps there, and how do you actually help companies to embrace some of these practices or concepts?
Marcio Saito: Let’s start with DevOps, which is the more generic term. DevOps is really collapsing development and operations, so that you have a more continuous cycle rather than you design the infrastructure, you deploy, and then you keep it untouched for 20 years, you try to follow a more continuous approach where you’re constantly designing, developing, deploying, getting feedback, and changing our design. That’s DevOps, it involves the application of automation, so it’s less about manual operations. Rather than you doing something, you should spend your time writing code that you can use to automate the process. This is in a nutshell, very high level, what DevOps is.
The same applies, we are trying to break the silos and make the management of infrastructure more uniform, so the DevOps principles also applies to NetOps. NetOps, network management people, they are trying to adopt the same thing, but a network is the most conservative of the silos, because first, they have more responsibility. If the network is down, everything is down, so they tend to be more conservative, so that’s one of the reasons why the network infrastructure is still not automated, is still not fully virtualized. But there is another aspect that makes NetOps or the network portion of DevOps different, which is again, the connection between the logical and the physical. You can virtualize, you can automate, but in networking, there is always the physical connection. You’re never going to get out of managing physical devices and mapping your logical resources to physical devices.
That’s why I use, and why Opengear uses the term NetOps because though DevOps applies to everything in IT, NetOps is still special, the network is still special because of the connection between the physical and logical layers. The networking people, they have different challenges when they try to automate because the automation has to work over physical devices. Again, it goes back to your first question, it’s very easy to scale compute nodes, is not as easy to scale networking devices on that network infrastructure. That’s why I do think DevOps is going to equalize all the other silos in IT, but networking is different and NetOps has to be seen as a separate discipline in the longterm because of the connection with the physical layer.